Share Email Print

Proceedings Paper

Removal of repeated objects in multi-channel images
Author(s): A. Connor; A. K. Forrest
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

This paper describes a technique for removing known repeated objects in multi-channel images. The motivation for removing known objects is to highlight un-known objects of interest. An important criterion is to base the technique on the use of image statistics and limit the use of operator supplied thresholds. The dependence on such thresholds may be subjective and is likely to be a source of error. The images used in this paper consist of transient eddy current data acquired from a multilayer aluminium alloy panel which represents a typical airframe structure. The image intensity represents the change in received eddy current magnetic field as a coil excited with a broadband waveform is scanned over the surface of the structure. The structure is held together using fasteners which are shown on the image as repeated circular objects. The regions around the fastener holes are of interest since this is where cracks in the structure may occur. Fastener objects have high variance in intensity compared to cracks and thus tend to dominate the image making identification of neighbouring cracks difficult. The removal of the known repeated objects (fasteners) highlights the un-known objects of interest (cracks). The data consists of 12 channels representing depth through the structure. Principal Value Decomposition (PVD) is to be used to reduce the number of channels that represent most of the information. The technique to remove known repeated objects involves computing the polar 2D FFT image frames. A circular object such as a fastener will produce a 2D Bessel function in frequency space. Since the object is circularly symmetric in frequency space the polar transform may be collapsed onto one axis representing radius. Zero crossings on this axis provide information on the inner and outer radius of the object. Larger frames produce grater noise from the many features present. Sub-frames can be used so that the algorithm will converge and the object radii may be measured. Correlation by multiplication in the frequency domain is used to determine the centre position of each fastener in the image with maximum signal to noise ratio. A function based on the physical process that creates the image of each fastener object is applied to the image to remove them. Two alternative techniques are presented. PVD can significantly reduce the number of channels that contain most of the information. The results show that the centre can be located and radii of each fastener can be measured, and the known image features removed to show subtle characteristics of the image not normally visible.

Paper Details

Date Published: 31 December 2008
PDF: 6 pages
Proc. SPIE 7130, Fourth International Symposium on Precision Mechanical Measurements, 71302X (31 December 2008); doi: 10.1117/12.819664
Show Author Affiliations
A. Connor, Imperial College London (United Kingdom)
A. K. Forrest, Imperial College London (United Kingdom)

Published in SPIE Proceedings Vol. 7130:
Fourth International Symposium on Precision Mechanical Measurements
Yetai Fei; Kuang-Chao Fan; Rongsheng Lu, Editor(s)

© SPIE. Terms of Use
Back to Top
Sign in to read the full article
Create a free SPIE account to get access to
premium articles and original research
Forgot your username?