Holographic particle image velocimetry apparatus and methods

Reference Number: K 98-21

Inventors: Meng, Hui; Pu, Ye

USPTO Link: 6496262

Invention Summary

It is an object of the present invention to provide a holographic particle imaging system that is capable of extracting three-dimensional detail from the flow field, including but not limited to individual particle sizes, shapes, displacements, velocities and concentrations.

Another object of the present invention is to provide a workable optical configuration for HPIV, which combines the simplicity of in situ reconstruction with high image quality and easy optical access. The present HPIV configuration's 90-degree scattering feature permits optical access approaching that of planar PIV, which has been widely accepted.

A further object of the present invention is to provide a system with high compression ratios for the data to be processed, together with an efficient velocity field extraction algorithm. The use of particle centroids for velocity field extraction, rather than the image data extracted from the hologram as is used in some HPIV systems, permits the use of less storage and active memory, which speeds the processing of the data. Furthermore, retaining the spatial locations of the particle centroids permits particle pairing to be done after extraction of the major features of the flow; clearly this provides the best possible resolution of the velocity vector field.

To achieve these objects, there is provided a system for the holographic recording of three-dimensional images of a fluid seeded with particles, and for the extraction of the velocity vector field defined by the motion of the fluid. The system includes the specification of an off-axis holographic system utilizing 90-degree scattering to obtain low depth of focus and high SNR, in situ reconstruction of the time-separated holograms, extraction of the particle centroid locations from the time-separated holograms, and a velocity field extraction algorithm which uses particle centroid data, rather than digital images captured from the hologram, to perform the correlations needed to calculate the three-dimensional velocity vector field.

The system utilizes an object beam and two reference beams that are directed onto the holographic recording medium from different reference angles. The different reference angles permit the particle fields recorded at different points of time, t1 and t2=t1+dt, to be unambiguously distinguished from each other. The object wave is obtained by directing a third coherent wavefront at an angle perpendicular to the vector between the particle field and the single recording medium, so that the 90-degree scattering component from this beam interferes with first one reference beam at time t1, and then the other reference beam at time t2=t1+dt. The use of two angularly separated reference waves permits the time-separated holograms to be recorded on a single holographic recording medium. This system geometry results in an improved Numerical Aperture and depth of focus, allowing three-dimensional information to be effectively extracted from the resulting holograms.

The complexity of the imaging system is somewhat ameliorated by the in situ reconstruction scheme, which uses substantially the same optics for reconstruction of the time-separated holograms. The two reference beams are recreated in exactly the same geometry used for recording; the holographic recording medium is placed in the same position with the emulsion facing in the opposite direction, so that the incident reference waves are effectively the phase-conjugates of the original reference waves. The object wave is not needed in the reconstruction process, and so is blocked. The reconstructed images are displayed alternately, on the opposite side of the holographic emulsion, by illuminating the emulsion alternately with the two reference beams.

The system then utilizes an imaging system which alternately interrogates the three-dimensional volumes reconstructed by the two reference beams. A CCD camera on a three-dimensional traverse system, with a small depth of focus, is and acquires planar images by scanning synchronized with the laser. The CCD camera alternately extracts intensity data from the two holograms, taking a thin slice in the depth dimension, and small areas in the planar dimension. This gives rise to a natural three-dimensional grid on each volume. The processor then employs standard particle centroid extraction algorithms to save the three-dimensional location of each particle centroid in the intensity map. The concise cross correlation algorithm (CCC) is then utilized to extract a velocity vector field defined on this grid.

CCC extracts velocity vector fields by measuring displacement in 2-dimensional and 3-dimensional media, including, but not limited to, fluid flows. In the preferred embodiment, particle centroid coordinates are extracted from each pair of images, or from image planes in the hologram. The coordinates alone are retained in the processing that follows, saving both compute time and memory. The first cube is naturally broken into target windows (cubes) by the volume interrogation process. These cubes are sufficiently small that their velocity fields are roughly constant at that scale. The correlation of each window (cube) with a region in the second image (cube) is calculated. The correlation is calculated on the basis of particle centroids alone, with correlation intensities between individual particles being calculated as a decreasing function of distance. Translations are calculated by manipulating centroid coordinates.

To accelerate computation, three 1-dimensional correlations are calculated in place of a complete 3-dimensional correlation. The velocity vectors are obtained by combining x-, y-, (z-) components obtained from the individual correlations. The highest possible spatial resolution can then be obtained, if desired, by translating the images by the extracted velocity and pairing particle centroids individually.

The entire computation can be performed in integer arithmetic only, which means that CCC can be implemented easily on hardware platforms for real-time applications. The method is robust, and much faster than the FFT-based methods in common use.