Detection and resolution of multiple targets using time-frequency and deconvolution techniques

Jianqiang Xin

doi:10.17918/00009821

In ultrasonic nondestructive testing (NDT) the backscattered echoes from the grain boundaries often mask the target signal, leading to difficulties in its detection and identification. The split spectrum processing (SSP) technique, which is based on the frequency diversity principle, has been used effectively to suppress grain noise when a single target is present. However, in general, more than one target may exist over the resolution cell, resulting in interference which can mask one or more of the targets and present difficulties in their detection and identification. It has been shown in this work that the inter-target interference between two targets exhibits strong variation with frequency. Therefore, by bandpass filtering the input signal at different center frequencies, it is possible to obtain an output signal which exhibits less interference between the adjacent targets compared to the original wideband signal. The simulation and experimental results show that by utilizing such bandpass signals, constant bandwidth and constant Q SSP algorithms can yield both signal-to-noise ratio enhancement (SNRE) and resolution improvement. Deconvolution techniques have been widely used to improve the resolution and quality of ultrasonic images. However, deconvolution techniques require a priori information about the system, which presents the main difficulty in their implementation. In this thesis, the Split Spectrum Processing (SSP) spectral histogram technique is utilized to estimate the system transfer function. The spectral histogram technique is based on the statistics of narrowband signals selected by the absolute-minimization operation. Theoretical analysis indicates that the spectral histogram is similar in nature to the Wiener filter transfer function and can therefore be used to estimate the optimal frequency region for L1 deconvolution. The theoretical and experimental results indicate that the proposed technique provides enhancement in identifying and extracting multiple targets. Finally, based on the SNR variation of the narrowband signals over the frequency range of SSP, an analysis is provided to determine the optimum rank for the order statistic filter with independent non-identically distributed inputs.

Detection and resolution of multiple targets using time-frequency and deconvolution techniques

Files and links (1)

Abstract

Metrics

Details

Detection and resolution of multiple targets using time-frequency and deconvolution techniques

Files and links (1)

Abstract

Metrics

Details

Drexel University Social media