Fast implementation of matched-filter-based automatic alignment image processing

Abstract

Video images of laser beams imprinted with distinguishable features are used for alignment of 192 laser beams at the National Ignition Facility (NIF). Algorithms designed to determine the position of these beams enable the control system to perform the task of alignment. Centroiding is a common approach used for determining the position of beams. However, real world beam images suffer from intensity fluctuation or other distortions which make such an approach susceptible to higher position measurement variability. Matched filtering used for identifying the beam position results in greater stability of position measurement compared to that obtained using the centroiding technique. However, this gain is achieved at the expense of extra processing time required for each beam image. In this work we explore the possibility of using a field programmable logic array (FPGA) to speed up these computations. The results indicate a performance improvement of 20 using the FPGA relative to a 3 GHz Pentium 4 processor.

Image processing and control of a programmable spatial light modulator for spatial beam shaping

“…Matched filtering (MF) [12,13] has demonstrated remarkable success in determining the location of distinctly shaped beam fiducials [14][15][16]. One of the chief advantages of MF is that it can be applied to the analog domain image without performing extensive preprocessing.…”

Image processing and control of a programmable spatial light modulator

“…Matched filtering (MF) [6,7] has been successfully utilized to determine the location of beam fiducials with distinct shapes [8,9]. One of the chief advantages of the technique is that MF can be applied to the analog domain image without performing extensive preprocessing.…”

“…The performance of the matched filter can be further enhanced by extracting the edge of the image and using the edge of the to-be-detected features as the filter. This has an equivalent effect of high-pass filtering the correlation output, thus increasing the sharpness of the peaks [9]. The position of the object can be found from the position of the cross correlation peak, autocorrelation peak, and the position of the templates.…”

New approach for the detection of noise‐distorted signals based on the method of S‐preparation

“…Developed with a preliminary processor of geometric images Requires prediction of the grey level of images. Increased computational intensity Optical matched filter [12] Analogue matched filter with the use of coherent processing of light Practically instant correction. Exceptionally big memory capacity.…”