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

Video images of laser beams imprinted with distinguishable features are used for alignment of 192 laser beams at the National Ignition Facility (NIF). Algorithms for determining the position of these beams enable control systems to perform the task of alignment. Real world beam images suffer from intensity fluctuation or other distortions, making algorithms susceptible to higher position measurement variability. Using matched filtering to identify beam positions results in greater stability of position measurement compared to centroiding techniques. However, this gain is achieved at the expense of extra processing time. This work explores the use of FPGAs to accelerate these computations. Results indicate a performance improvement of 20 times for an FPGA over a 3 GHz Pentium 4 processor.