Real-time full matrix capture for ultrasonic non-destructive testing with acceleration of post-processing through graphic hardware

Fast total focusing method for ultrasonic imaging

“…This point is the main issue of TFM, which is difficult to apply in real-time applications. In recent years, works have been focused on parallelization through GPP or GPU [7]. However, the larger number of sums enables the reduction of the insignificant signals such as noise, which increases the signal to noise ratio.…”

“…The main issue is definitely the computational cost of the reconstruction algorithms. According to the output image size, the computation time can be prohibitive, in particular for real-time applications [7]. The migration approach has been introduced in the geophysics community by Stolt in the 1970's [8].…”

Full matrix capture with time-efficient auto-focusing of unknown geometry through dual-layered media

“…For the use of FMC data, the application of a look-up table has been shown to provide a significant performance increase by reducing the number of calculations in favour of data retrieval operations [8] , but only at the cost of increased memory requirements. For modern 64-bit computers this is less of an issue as much more memory is addressable by an application than may previously have been available.…”

“…This would require a look-up table of dimensions x, z, tx, rx providing the storage requirements for the necessary performance boost. There would still remain an initial greater computationally expensive task as this look-up table is first populated, but for all subsequent operations retrieval of this pre-calculated data has been shown to be far quicker than without [8] ; a technique that would work well for inspections normal to the surface or where a fixed perspex shoe is used (as in the case of angle wedge inspection).…”

“…Breaking down Equations (1) and (3) into smaller components allows for the removal of any dependency existing between operations, making the algorithm a suitable candidate for execution over the GPU, which has been shown for a planar interface [8] . Figure 3 illustrates this process for a non-planar surface where all time-critical operations are performed on the graphics hardware.…”