GB2510862A

GB2510862A - Image velocimetry

Info

Publication number: GB2510862A
Application number: GB201302672A
Authority: GB
Inventors: David James Cleaver
Original assignee: University of Bath
Current assignee: University of Bath
Priority date: 2013-02-15
Filing date: 2013-02-15
Publication date: 2014-08-20
Also published as: WO2014125261A9; WO2014125261A1; EP2956779A1; GB201302672D0

Abstract

In DDPIV (Defocused Digital Particle Image Velocimetry) measurements, colour may be used to give multiple parameters (e.g. R-G-B intensity) instead of one (monochrome intensity) for correlation and particle tracking. This is achieved using laser induced fluorescence applied to tracking particles 206. These particles are illuminated by laser light and fluoresce in different colours allowing individual particles to be differentiated more easily. Camera units 202, pulsed laser 110 and a timing unit 120 are used to capture time-separated samples of liquid flow in a target volume and to produce related colour images with component intensity samples. Each pulse has a predetermined peak frequency and the fluorescent tracking particles have an absorption frequency band including the predetermined peak frequency. Seeding ratio detection (seeding different phases of a multi-phase flow with different colours), and/or temperature detection (through seeding with thermo-liquid crystals) may also be used.

Description

IMAGE VELOCIMETRY

The present invention relates to apparatus and methods for measuring the velocity of flow fields in fluids. Embodiments of the invention relate to the measurement of velocity vectors, typically decomposed into three orthogonal velocity components, in a three dimensional volume of the fluid.

BACKGROUND OF THE INVENTION

Within the field of experimental fluid dynamics there is the demand to measure the velocity of flow fields in fluids, however this is inherently difficult due to the homogeneity of the fluid (air, water, etc.). One of the features of this homogeneity is the uniformity of the opfical properties of the fluid; many fluids of interest are uniformly optically transparent or translucent. To measure the velocity of flow fields, then, the fluid is typically seeded with small particles (e.g. hollow glass spheres in water; smoke in air, etc.). These particles, sometimes referred to as "tracer" or "tracking" particles, need to be small enough to follow the fluid flow (without significantly affecting that flow) and substantially neutrally buoyant, yet capable of being discriminated within a measurement volume. The motion of the fluid can then be indirectly measured through the motion of the tracking particles using a variety of techniques such as: Laser Doppler Velocimetry (LDV): single point measurement using Doppler shift of particle passing through two / three collimated laser beams.

ii. Particle Image Velocimetry PIV): 2D (two velocity components, u and v) planar measurement using a laser sheet to illuminate the particles and a monochrome camera to record the image. A second image is acquired a small time later. The motion of the particles is measured by dividing the image into interrogation windows (e.g. 16 x 16 pixels) and two-dimensional mathematical cross correlation performed to calculate the motion of the interrogation windows. An alternative, but much less common form of Ply, is particle tracking velocimetry (PTV) where individual tracer particles are tracked rather than the motion of interrogation windows.

iii. Stereographic FlY: a Ply technique using two cameras to give tIn-cc velocity components (u,v,w).

iv. Defocused Digital Particle Image Velocimetrv (DDPIV): (also known as Digital Particle Tracking Velocimetry (DPTV)) uses three or more cameras to track parficles and thereby derive three velocity components (u,v,w) in a (three dimensional) volume.

DDPIV is described in greater detail below.

v. Tomographic PlY: uses a plurality of cameras to derive a three-dimensional model of the volume and monitor the deformation', to give three velocity components (u,v,w) in a 3D volume.

The ultimate goal of these techniques is to acquire instantaneous velocity fields accurately. However, due to inherent limitations in the respective techniques (e.g. LDV can only measure one point at a time) or due to poor data quality (for DDPIV and Tomographic PIV) it is usually necessary to average the measurements instead.

US patents, US 7006132 and US 6278847, describe camera arrangements suitable for use in the DDPIV tcclrnique. An image detector is provided with an aperture mask and light from each aperture is directed by a lens arrangement either onto a plurality of cameras or a single camera (the light from each aperture being simultaneously collected in a different portion of the photographic or CCD camera). In one embodiment, a prism arrangement is used to direct light fiom apertures in the aperture mask to corresponding monochrome cameras.

Another approach to image velocimetry that has been considered is based upon a combination of Ply with Planar Laser Induced Fluorescence (PLIF). Laser Induced Fluorescence (LIF) is a phenomenon iii which incident laser light (i.e. photons at a frequency f1) is absorbed by particles of certain chemical species (for simplicity, these are referred to as "dyes"), which in turn re-emits some of the energy thus absorbed as fluorescence. To give fluorescence, the absorbed energy must be sufficient to raise a significant number of electrons of the dye particles to an (unstable) excitation state; i.e. the frequency of the incident photons corresponds to a frequency band at which the dye particles resonate. The excited dye particle then spontaneously re-emits some of the absorbed energy (i.e. photons at a frequency f2, where f2<f). The dye species is lypically an organic fluorescent dye such as fiuoreseein or rhodamine dye. The emission band from any one dye species is generally narrow, to sonic extent mirroring the absorption band, giving the fluorescence a distinctive colour.

PLIF is a method suitable for measuring the phase density in multi-phase flows using LIF. Here the term "multi-phase" refers to two related concepts: either the input of the same fluid from multiple sources in mixing studies; or the input of distinct fluids from one or more sources. The different flows/fluids are seeded with dye particles which exhibit LIF of respective different colours and the ratio of these particles (calibrated) determines the phase ratio. For effective PUP, laser and dye particle species need to be considered as a pair so that the laser delivers power in at least part of the resonance band of the dye particles. Where different dye particles are used they must be chosen so that they each have resonance bands that overlap the (narrow) frequency band of the laser. It is noteworthy that until recently the narrow emission band of dye particles had discouraged use of colour cameras to capture emission images, monochrome camera systems can give improved accuracy without the additional cost or complexity.

To measure the velocity of flow fields, hybrid PLIF-PIV systems are known and at least one researcher has described attempts to use colour cameras simultaneously to acquire PLIF and Ply measurements. The colour cameras used had Bayer filter format and, as a result, the particles were too indistinct for effective PlY measurements, and this technique was discarded. Where both PLIF and PIV measurements are needed, separate monochrome (Ply) and colour (PLIF) camera arrangements are collocated.

It is also known to combine the [IV technique with Theniiotropic Liquid Crystals (TLC). Much as the LIF dye particles are differentiable by virtue of their fluorescence properties, TLC particles react to the temperature of the fluid around them by changing their colour.

SUMMARY OF THE INVENTION

According to a first aspect of the invention there is provided an apparatus for capturing time-separated image samples of fluid flow in a target volume, the volume being seeded with a plurality of tracking particles, the apparatus comprising: a pulsed light source, which operates to deliver a first pulse of light at a first time and a second pulse of light at a second time, each pulse of light having a predetermined peak frequency; a camera arrangement including a plurality of colour camera units, which in operation capture sets of component intensity samples by capturing images oF the tracking particles in the target volume; a timing unit for synchronising the camera arrangement to capture the respective scts of component intensity samples with thc delivery of the respective pulses of light capture, a first set of component intensity samples being captured at the first time and a second set of component intensity samples being captured at the second time; wherein the tracking particles are fluorescent type tracking particles having an absorption frequency band substantially including the predetennined peak frequency, and wherein the camera units each generate respective colour images, each colour image comprising a plurality of component intensity samples.

The pulsed light source is preferably a laser source.

According to another aspect of the invention, there is provided a method for capturing time-separated image samples of fluid flow in a target volume, the volume being seeded with a plurality of tracking particles, the method comprising: delivering a first pulse of light at a first time and a second pulse of light at a second time, each pulse of light having a predetermined peak frequency; capturing, in a plurality of colour camera units, a first set of component intensity samples corresponding to images of the tracking particles in the target volume while being illuminated by the first pulse of light and a second set of component intensity samples corresponding to images of the tracking particles in the target volume while being illuminated by the second pulse of light, the capture of the respective sets of component intensity samples being synchronised with the delivery of the respective pulses of light capture, wherein the tracking particles are fluorescent type tracking particles having an absorption frequency band substantially including the predetermined peak frequency and the camera units each geueratc respcctive colour images, each colour image comprising a plurality of componcnt intensity samples.

According to a furthcr aspect of the invention thcrc is providcd method for determining velocity componcnts of tracking particlcs in a targct volumc, thc tracking particles being fluorescent type tracking particles having individual colour characteristics, the method comprising: generating a first set of component intensity samples at a first time; identifying discrete particles in each component intensity sample in the first set; determining a correspondence between a discrete particle identified in a first component intensity sample of the first set and a discrete particle identified in a second component intensity sample of the first set, said correspondence being dependent upon the colour characteristics of the discrete particles identified in the first set of component intensity samples, generating a second set of component intensity samples at a second time; identifying discrete particles iii each component intensity sample in the second set; determining a further correspondence between a discrete particle identified in a third component intensity sample of the second set and a discrete particle identified in a fourth component intensity sample of the second set, each further correspondence being dependent upon the colour characteristics of the discrete particles identified in the first set of component intensity samples, matching particles for which correspondence is determined in the first set to particles for which correspondence is determined in the second set, each match being dependent upon the colour characteristics of the correspondence particles; and determining the change in location of each of the matched particles for which a matching particle is found in the second set of samples, thereby determining the velocity components for the respective particles.

In accordance with yet another aspect of the present invention there is provided an apparatus for determining velocity components of tracking particles hi a target volume, the tracking particles being fluorescent type tracking particles having individual colour characteristics, the apparatus comprising: means for generating a first set of component intensity samples at a first time; means for identifying discrete particles hi each component intensity sample in the first set; means for determining a correspondence between a discrete particle identified in a first component intensity sample of the first set and a discrete particle identified in a second component intensity sample of the first set, said correspondence being dependent upon the colour characteristics of the discrete particles identified in the first set of component intensity samples; means for generating a second set of component intensity samples at a second time; means for identifying discrete particles in each component intensity sample in the second set; means for determining a further correspondence between a discrete particle identified hi a third component intensity sample of the second set and a discrete particle identified in a fourth component intensity sample of the second set, each further correspondence being dependent upon the colour characteristics of the discrete particles identified in the first set of component intensity samples; means for matching particles for which correspondence is determined in the first set to particics for which corrcspondcnce is determincd in the sccond set, each match being dependent upon the colour characteristics of the correspondence particles; and means for determining the change in location of each of the matched particles for which a matching particle is found in the second set of samples, thereby determining the velocity components for the respective particles.

Various aspects and features of the present invention are defined in the appended claims.

It will be appreciated that features and aspects of the invention described above in relation to the first and other aspccts of the invention are equally applicable and may be combined with embodiments of the invention according to the different aspects of the invention as appropriate, and not just in the specific combinations described above.

BRIEF DESCRIPTION OF DRAWINGS

Embodimcnts of the present invention will now be described by way of example only with reference to the accompanying drawings where like parts arc provided with corrcsponding reference numcrals and in which: Figure I provides a schematic diagram illustrating an architecture ol an image detector system used in conventional DDPTV image velocimeiry; Figure 2 provides a schematic diagram illustrating an architecture of an image detector system used in the DDPIV image velocimetry technique of the present invention; Figures 3A to 3C show the key stages in the image velocimetry technique of the present invention; Figure 4 illustrates three camera sensor formats suitable for use in colour camera units; and Figure 5 illustrates a further embodiment of the image velocimetry technique of the present invention

DETAILED DESCRIPTION

Figure 1 illustrates an image velocimefty arrangcment. A camera arrangement 102 (here comprising three camera units 102-1, 102-2, 102-3) is directed towards a target volume 104 of optically transparent/translucent fluid. The fluid to be observed is seeded with tracking particles (tracer particles) 106. The fluid may either be premixed with a suitable quantity of these particles or a tracer source may be provided at a point "upstream" of the target volume 104 and arranged to diffuse the particles in the fluid so that they are substantially homogenously distributed throughout the fluid as it enters the target volume. The cameras need to be precisely positioned to ensure that they each target substantially the same volume, and are therefore capable of capturing images of the same particles.

A light source, typically a laser 110, illuminates the tracking particles 106. The laser used to illuminate the particles is a conventional pulsed laser (such as a Nd:YAG laser) with appropriate optics (not shown), and a timing unit 120 (also known as a "synchronizer") to coordinate timing betwcen the laser 110 and camera units 102-1, 102-2, 102-3. Laser illumination is necessary to provide adequate illumination of the target particles in a sufficiently short time interval that the tracking particle will not be displaced significantly over the period of image capture: in other words, the pulse duration is much shorter than the time taken by the average tracking particle to cross the target volume. As a result, the duration is short enough that the particle is effectively "frozen" in space, thereby substantially removing motion blur in image capture. References to "instants of time" thus in fact refer to very short periods of time (of the order of nanoseconds) over which no significant movement of the particles can be discerned.

In a typical DDPIV apparatus, three monochrome camera units are arranged to capture respective sets of images of laser-illuminated tracking particles within a three-dimensional volume at two instants of time, t1 & t2. The capture images are transferred to a computing device 130; typically a general purpose computing device 130, such as a desktop personal computer (PC:) or laptop PC, but occasionally, a computing device dedicated to the processing of image data.

In addition to having interfaces with the camera arrangement 102 and the timing unit 120, the computing device 130 includes a processing unit 132 and a storage unit 134. First images, imagei.i image1.2 and image1,3, captured by the respective camera units 102-1,102-2,102-3 at the first time instant, ti, are stored in the storage unit.

The processing unit then identifies a plurality of particles within the first images, imagei,j. In the three first images, the pixel locations of each particle are determined: provided the particle fulfils a selection criterion (e.g. peak intensity above a predetennined threshold, particle radius greater than a predetermined size, etc.), the pixel location of the candidate particle is stored. Comparing the particle locations in at least two of the first images, identified particles at similar pixel locations are considered to be the same particle (with differences in intensity and location being attributed to the different points of view of the respective camera units): the respective identified particles in each image are said to be "paired".

Tracking particles are thus identified in each of the three first images captured at the first time instant, t1: a step referred to as the discrete particle detection step. The pairing of identified particles, through calibration between the three images, gives coordinates, x1 = (x,y,z), of the paired particle within the volume at the first time instant, t1.

Second images, image2,1 image2,2 and image2,3, captured by the respective camera units 102-1,102- 2,102-3 at the second time instant, t2, are stored in the storage unit. As for the first images, the processing unit identifies particles within the second images, image2,i. In the three second images, the pixel locations of each particle are determined: provided the particle fulfils the selecfion criterion (i.e. the same identification criterion as for the first images), the pixel location of the candidate particle is stored. Particle locations in at least two of the second images are compared and, again, identified particles at similar pixel locations are considered to be the same particle.

Tracking particles are thus identified in each of the three second images capturcd at the second time instant, t2. The pairing of identified particles, through calibration between the three images, gives coordinates, = (x,y,z), of the paired particle within the volume at the second time instant, t2.

In other words, the steps of particle identification and pairing are performed for a first set of (three) images and repeated for a second set of (three) images to give two sets of particle coordinates representing two respective time instants, t1 & t2.

The particles (characterised by their coordinates) are then matched between the two times using the similarity of [heir relative distance horn neighbouring particles; in a process referred to as [he particle matching step or "near neighbour" analysis. This gives their displacement in the known time (t2 -t1) and thus their velocity v = (u,v,w).

There are a number of inconvenient features of the known techniques. DDPIV (DDPTV) techniques suffer fi'om ambiguity in the particle pairing and particle matching steps. It can be difficult to distinguish between individual particles within each set of images. Within the particle pairing stage all particles are seen as identical and therefore at least three cameras are required (in effect two to pair and the third to validate). Similarly, within the particle matching stage, the particles all appear as identical with the only distinguishing feature being the distance to neighbouring particles.

Furthermore, DDPIV / DDPTV techniques typically require relatively high seeding densities (using of the order 40 particles in a target volume of 1cm3) due to the neighbourhood requirement in the particle matching step (matching between two time instants). Naturally, the particles need to be present in sufficient densities for there to be a statistically significant number of neighbours in close proximity otherwise the distance to neighbouring particles will change between ti and t2 (due to local variations in velocity within the volume), resulting in failure at the particle matching stage. Due to this minimum seeding requirement DDPIV is generally not practical in air.

On the other hand, the presence of too many tracking particles, referred to as "image crowding", leads to obscuring of particles and thus prevents accurate detection, pairing and matching. A preferred number of neighbour particles within a "cell" uscd in thc particle matching step can bc anywhcrc between 3 and 30. Empirically, having 8 neighbour particles on average in a cell appears most effective. Assuming the seeding density is the same, using fewer than 8 neighbour particles results in an increase ratc of failed matching (and therefore background noise), whilc using progressively more than 8 ncighbour particles mcans that the particles will inevitably be further apart resulting in movement relative to each other between t1 and 12, (again giving rise to increased noise), ftirtherniore the computational overhead (e.g. load, time, etc.) also increases. [-lowever, computational time is an inconvenience rather than a limitation because the results are typically processed later (not in real time).

In Tomographic Ply, too, multiple monochrome cameras capture the volume of interest. A three-dimensional model is built of the volume at two time instants and the volume is effectively tTeated as a deforming liquid (analogous to three-dimensional [IV).

In one embodiment of the invention, laser induced fluorescence (LIP) is applied to tracking particles so as to shift the wavelength of the laser light scattered from the particles. The wavelength of the light scattered is ideally entirely random (within a detectable continuum) to give the widest possible range of colours and thereby aid discrimination between particles: in practical terms, it is adequate to provide tracking particles dyed to produce LIF scattering in a finite number of bands with minor variability due to production variability.

This improves the data quality of DDPIV measurements through the use of colour to give multiple parameters (e.g. R-G-B intensity, or luminosity Y and colour difference Cb/Cr) instead of one (monochrome intensity) for the correlation / particle tracking steps. In addition to wavelengths in the visible spectrum, the term "colour" should be understood to relate to wavelengths in the infrared or indeed in the ultraviolet, thus wavelengths outside the spectrum visible to the unaided human eye.

While there may be practical difficulties in achieving a LIP which induces a wavelength shift into certain parts of the colour spectrum, the ability to access a wide range of colours allows for a greater range of possible intensity/luminosity/colour difference parameters.

While this teelmique primarily improves upon the DDPIV technique (due to the difficulty of pairing particles in three-dimensional space), a similar method may also be applied to the Tomographic PIV and [IV techniques in general.

In addition, it is feasible to combine with seeding ratio detection (seeding different phases of a multi-phase flow with different colours), or temperature detection (through seeding with thermo-liquid crystals).

Figure 2 illustrates an image velocimetry detector system in accordance with embodiments of the invention. Where elements are substantially the same as those in the conventional DDPIV detector system illustrated in Figure 1, like reference signs relate to like components.

A camera arrangement 202 (here comprising three camera units 202-1, 202-2, 202-3) is directed towards a target volumc 104 of optically transparent/translucent fluid. Here, the fluid to bc observed is seeded with LIF tracking particlcs 206.

Again, a laser 110 illuminates the tracking particles 206 and as previously explained, the laser used to illuminate the particles is a conventional pulsed laser (such as a Nd:YAG laser) with appropriate optics (not shown), and a timing unit 120 (also known as a "synchronizer") to coordinate timing between the laser 110 and camera arrangcment 202.

As for conventional DDPIV, laser illumination is necessary to provide adequate illumination of the target particles in a sufficiently short time interval that the tracking particle will not be displaced significantly over the period of image capture. The laser 110 serves a ftirther purpose in this case as the laser is arranged to stimulate fluorescence from the LIF tracking particles (i.e. the frequency of operation of the laser is chosen to lie within the absorption frequency band of thc LIF dyc species).

As indicated, the camera arrangement 202 need only include two camera units 202-1, 202-2; a third camera unit 202-3 is optional. Whereas the camera units in Figure 1 are monochrome, the camera units 202-1, 202-2 (and optionally 202-3) in camera anangcmcnt 202 are colour camera units.

Typically each colour camera unit captures images as a pixelated array of intensity components, so that each pixel has an associated plurality of intensity components, for example intensity components corresponding to a point in RGB colour space. The colour camera units are arranged to capture respective sets of colour images of laser-illuminated LIF tracking particles 206 within a three-dimensional volume 104 at two instants of time, ti & t2. The capture images are transferred to a computing device 130 for processing and storage.

In addition to having interfaces with the camera arrangement 202 and the timing unit 120, the computing device 130 includes a processing unit 132 and a storage unit 134. First colour images, colour_image1,1 and colour_image1.2 (and optionally colour_image1.3), captured by the respective camera units 202-1,202-2 (and optionally 202-3) at the first time instant, t1, are stored in the storage unit 134.

Particle identification (discrete particle detection) proceeds much as in the preceding case. This is illustrated in Figure 3A. Here three first colour images (305-1, 305-2 and 305-3) are captured at time, ti, andthree secondcolourimages (315-1, 315-2 and3lS-3) arecapturedattime,t2.

The processing unit 132 identifies a plurality of particles within the first colour images (305-1, 305-2 and 305-3), colour imagei. For simplicity, Figure 3A shows the identification of a single tracking particle 310. This tracking particle appears in each of the first colour images (305-1, 305-2 and 305-3) as 310-1, 310-2 and 310-3 respectively, the pixel locations of each appearance 310-1, 310-2 and 310-3 S of the particle are determined and provided the respective particle appearances fulfil a selection criterion (e.g. intensity above a predetermined threshold), the pixel locations of the candidate particle appearances are stored.

The characteristic fluorescence colour values of the individual tracking particles assist in particle pairing (shown in Figure 3U). In Figures 3A to 3C, for example, the tracking particle 310 has a strong green intensity component. In a one pairing method, the locations for appearances 310-1, 310-2 etc. of the tracking particle 310 in at least two of the first colour images, are used to decide on the pairing between images: identified particles at similar pixel locations are considered to be the same particle 325 (with small differences in intensity and location being attributed to the different points of view of the respective camera units). Pairing of these appearances is thus preferably perfoniied within a degree of tolerance of differences in angular and/or pixel displacement 335. In this method the detected colour (in terms of RGB values for instance) of the identified particle, is then used to validate the pairing -differently coloured particles in different images taken at the same time instant cannot be considered to be the same particle.

In an alternative method, the detected colour (in terms of RGB values for instance) of the identified particles in each of the first colour images is determined first. An initial pairing can be generated by finding tracking particles of matching colour: appearances of identified particles of substantially the same colour in at least two of the first colour images are considered to be candidate appearances of the same particle. Only then are the locations, for appearances 310-1, 310-2 etc. of tracking particles 310 of substantially the same colour in at least two of the first colour images, used to decide on the pairing between images. The similarity of the pixel locations is considered validation that the same-coloured particle appearances are really the same particle. In this alternative method, the calibration of particles of similar locations in the first colour images is used to validate the initial pairing inferred from the detected colour of the tracking particle particles identified in respective first colour images.

Tracking particles are thus identified in each of the first colour images captured at the first time instant, t1. The pairing of identified particlcs, through calibration between the three images, gives coordinates, x1 = (xi,y1,zt), of the paired particle within the volume at the first time instant, ti.

Second colour images, colour_image2,1 315-1 and colour_image2.2 315-2 (and colour_image2.3 315-3), captured by the respective camera units 202-1,202-2 (202-3) at the second time instant, t2, are stored in the storage unit 134. As for the first colour images 305, the processing unit 132 identifies appearances of particles within the second colour images, colour_image2. In the second colour images 315, the pixel locations of each appearance are detenmned: provided the particle fulfils the selection criterion (i.e. the same identification criterion as for the first colour images), the pixel location of the candidate particle is stored.

Pairing of particle appearances in the second colour images (315-1, 315-2 and 315-3) can be performed in either of the alternative methods described above: with calibration of appearance locations followed by validation by colour value or initial pairing of particle by colour followed by confinnation of the pairing by location calibration.

Tracking particles are thus identified in each of the second colour images captured at the second time instant, t,. The pairing of identified particles, through calibration between the three second colour images, gives coordinates, X2 = (x,,y2,z2), of the paired particle within the volume at the second time instant, t2.

In other words, the steps of particle identification and pairing are performed for a first set of two (three) colour images and repeated for a second set of two (three) colour images to give two sets of particle coordinates representing two respective time instants, ti & t2.

The colour values of the individual tracking particles also assist in particle matching between the two time instants (shown in Figure 3C). Assuming the tracking particle 310 (with its strong green intensity component) appears iii images captured at both time instants, ti & h, the presence ola tracking particle having substantially the same colour intensity components in each set of images is used to match those two particles locations (characterised by respective coordinates).

Alternatively, a conventional "near neighbour" analysis may be applied with the colour value being used to confirm the particle match. The pai-ticles (charactcrised by their coordinates x1 and at respective times) are then matched between the two times using the similarity of their relative distance from neighbouring particles. This gives their displacement in the known time (t2 -t1) and thus their velocity v = (u,v,w). Again, the matching between candidate particles can be made significantly more efficient because it can be assumed that each particle has a characteristic colour which will be detected in images for each time instant. Figure 3C shows the displacement of tracking particle 310 between the two time instants: T1(R15, G240, 132) and T2 (RIO, G242, BO) representing their coordinates and the associated intensity components, in brackets. As the reader will appreciate, the intensity components at the two times are substantially the same (tracking particle 310 appears to be a bright green in both instances).

While only improvements to DDPIV are considered in detail, it will be clear to the skilled reader that LIF tracking particles coupled with suitable colour camera units may be applied more broadly.

II

Furthermore, for simplicity the system is desenbed in terms of three colour cameras that produce RGU format images. However, this does not preclude the use of a different number of cameras or indeed other cameras for obtaining images in alternative image formats (such as YCb/Cr, XYZ, etc.).

In the present invention, as illustrated in Figures 3A to 3G. coloured tracking particles are used not to distinguish between different flows or phases of flow but rather to provide an additional feature of each discretc tracking particle which can be uscd to detect that particle with greater certainty. By using colour cameras to capture images with a plurality of intensity components, individual particles may be identified more reliably in both the pairing and matching stages. Using multiple intensity components (e.g. R-G-B) as opposed to a single intensity (as has been the case with the monochrome cameras uscd to datc) significantly improves the accuracy of the DDPIV / DDPTV techniques.

The availability of multiple intensity components means it is possible to lower the seeding density.

This reduction can be dramatic: in the case where the colour parameter is used directly to pair and match particles in the images captured at t1 and t2, there need only be a single particle in the whole volume (effectively particle tracking completely removes the neighbour particle requirement and will therefore work with zero neighbours).

Even where neighbour displacements are used in line with the conventional DDPIV techniques, colour adds an extra criterion. The fidelity is significantly improved giving comparable results for seeding densities giving far fewer neighbour particles (the preferred number of neighbour particles being 2 or 3 as opposed to 8).

Furthermore, it will be recalled that, in conventional techniques, a third camera is used essentially as validation in the particle pairing stage. By providing tracer particles which can be discriminated on the basis of their colour properties, this third camera becomes superfluous (or at least optional) -colour may be used as validation instead, meaning that only two cameras would be required.

In the discussion of monochrome and colour DDPW techniques above, three camera units 102,202 are conveniently arranged in an equilateral triangle with a tilt inwards (see Figures 1 and 2). The camera units may be selected from a group of possible sensor formats including the three formats illustrated in Figure 4: Bayer filter 401, silicone sensor stack 402, and three separate GCD with trichroic prism assembly 403.

The Bayer filter 401 is described by the eponymous inventor in US patent 3,971,065. A mosaic layer of colour filters (typically G-green, R-red, and B-blue) is superposed over a photosensor array. The filter "tiles" may be placed in different patterns which in turn are refened to by the characteristic repeat sequences of the respective filters RGBG, GRGB, or RGGB etc. In the silicone sensor arrangement 402, such as that used in the Foveon X3 sensor stack, light of different wavelengths are absorbed at sensor layers at different depths in a silicone sensor stack. The silicone sensor conveniently allows the acquisition of wavelength information at a plurality of wavelengths simultaneously within a single sensor.

The three CCD format 403 could achieve similar performance to that of the silicon scnsor. Here ro dichroic prisms are assembled to form a "trichroic prism" which is used to split a single image into red, green, and blue components so they can be separately detected on three CCD arrays.

As noted above, pairing can be done on the basis of colour instead of (or in addition to) neighbour displacements allowing for highcr data quality with far fewer seeding particles. This is particularly advantageous in air where (to date) it has not been practical to achieve sufficient seeding density.

In an ahernative cmbodiment of the present invention, illustrated in Figure 5, data acquisition is implemented at a higher frame rate. The precise frame rate depends upon a wide range of factors: flow velocity, seeding density, volume size, camera resolution and so forth. Typical values would be on the order of 100Hz in water, and on the order or 1 kHz in air. As a result thc time between data sets is low enough to be able to perfonu the particle matching continuously. In other words, the matching is performed between successive pairs of image sets {t1,t2}, {t2,t}, {t3,t4}, etc. rather than in discrete pairs of image sets t1,t0}, {t1,t4} etc. The light source is again conveniently a laser (again this is to deliver a sufficient intensity in a short pulse). To achieve minimal power variation across the volume for the period of continuous operation, some minor adjustment of the laser apparatus may be necessary. The resulting, evenly distributed, pulsed light source is directed (as before) onto the target volume seeded with tracking particles (which have been prepared so as to exhibit LIF in a range of distinct colours). The system is thus suitable for tracking distinct particles throughout their transition through the target volume. In effect, the number of sets of images captures by the system can be increased beyond the two of the DDPIV technique described in relation to Figure 3A to 3C. As a result, the system may give true' streamlines which in conventional techniques arc more usually extracted through interpolation.

Using the same cameras as previous embodiments, it is thus possible to acquire seeding ratios simultaneously for multi-phase flows. In a ifirther embodiment of the invention, each flow is seeded with particles of a distinct fluorescent colour (or a colour selected from a distinct portion of a continuous range of colours). However, the use of a small number of discrete colours would inevitably degrade the efficacy of the colour-based particle tracking methods.

Thermotropic liquid crystal coated tracer particles may be used in addition to, or as an alternative to, the coloured LIF particles of previous embodiments. Where TLC tracer particles are used, it becomes possible to use the same cameras to acquire temperature and velocity information, with the same "trade off" between added ifinctionality and degradation in the efficacy of the particle tracking methods.

Embodiments of the invention differ from conventional systems in the following respects, at least: Laser Induced Fluorescent (LIF) particles for velocimetry. Previously LIF has been used for the PLIF technique where the fluorescence demarks the phases in a multi-phase flow, but never in velociinetry.

Colour camera for particle velocimetry. Previously colour cameras have been unsuccessftilly experimented with for a hybrid PUP-Ply system, see above. However, due to the development of new sensors (e.g. the silicon based Foveon X3), and the possibility of a threc-CCD system (see Figure 2) a single colour camera is now feasible.

Calibration. In terms of hardware, the calibration of the detector equipment can be performed through a variety of methods (e.g. transiting an x-y calibration plate through the z-plane, translucent block with particles fixed in regular x-y-z locations etc.), in all eases the common feature is a regular known grid of small distinct points placed throughout the volume of interest. These are illuminated with a white light source (to give high intensity values across the entire colour spectrum), and each colour within each sensor calibrated separately. This will produce 9 calibration curves (3 cameras x 3 colours), as opposed to the three from the monochrome camera arrangement used in conventional DDPIV. This calibration will serve both for the particle pairing stage (discussed in relation to Figure 3B) and for aligument between the different colour sensors (if necessary).

Image-particle detection. First the particles must be located in the six individual images (three cameras for two time instants) to give image-particles (in pixel locations), see Figure 3A. The current monochrome DDPIV system uses identification criteria such as peak intensity and particle radius. For colour DDPIV, depending on the sensor format (Bayer, silicon, three-CCD) andior colour filter pattern (RGBG, RGGB, CYYM...) used, the image data could be in a range of formats. In all cases there will be multiple colour matrices for multiple cameras. The particle identification can be performed through two possible methods: 1) Individual colours -using intensity criteria locate the image-particles in the different colours separately, if the intensity requirements are met in a single colour this is valid; 2) Combined colours -multiply the colour matrices together to produce a single matrix for each camera, perform intensity detection as before. Once the image-particles are located, the particle is given an image-location and a colour-mix', for example in RGB fonnat the values could be: (X200, Y250, R256 G50 BO).

Particle pairing. The image-particles must now be paired in the three separate images to give triplets which equate to: particle locations (in x-y-z) and particle colours (e.g. RGB) for the two time instants.

The current monochrome DDPIV system uses calibration triangles. In the calibration triangle technique, a line is drawn in image 2 from the current from image I image-particle along the vector defined in the calibration, any particle that dissects the line (within a tolerance) is a possible, a frirther line is drawn along the calibration-vector in image 3 to find another possible (within a tolerance), if valid the three image-particles will therefore form the vertices of a triangle with the calibration vectors forming the edges, see 335 in Figure 3B for an example. The centre of this triplet defines the x-y coordinate and its size the z-coordinate. For colour DDPIV the triplet stage pairs the particle-images between the three cameras using both pixel locations and colour intensity components. As is explained above, two possibilities for this stage are: i) calibration vectors used for triplet identification as before, colour-mix acts as validation; or alternatively, ii) colour-mix used for identification, calibration vectors used as validation and to give x-y-z location.

Velocity pairing. Now the particles have been identified in terms of x-y-z locations in the two times, they must be paired between the times to give the distance moved. The current monochrome DDPIV method uses the distance to neighbouring particles (typically eight) in an iterative process. There are two possible methods for colour DDPIV: i) to use relative displacements as before ["near neighbour" analysis], but add colour-mix as a validation criterion; or alternatively ii) to locate the same particle based on nearest colour-mix alone (within certain criteria: tolerance on difference in colour-mix, maximum particle displacement, particle not already allocated). For the latter method, it may be necessary to perform this in an iterative cycle to prevent multiple allocations of a single particle and thus refine the solution.

Claims

CLAIMSI. An apparatus for captunng time-separated image samples of fluid flow in a target volume, the volume being seeded with a plurality of tracking particles, the apparatus comprising: a pulsed light source, which operates to deliver a first pulse of light at a first time and a second pulse of light at a second time, each pulse of light having a predetermined peak frequency; a camera arrangement including a plurality of colour camera units, which in operation capture sets of component intensity samples by capturing images of the tracking particles in the target volumc; a timing unit for synchronising the camera arrangcment to capture the respective sets of component intensity samples with the delivery of the respective pulses of light capture, a first set of component intensity samples being captured at the first time and a second set of component intensity samples being captured at the second time; wherein the tracking particles are fluorescent type tracking particles having an absorption frequency band substantially including the predetermined peak frequency, and wherein the camera units each generate respective colour images, each colour image comprising a plurality of component intensity samples.
2. An apparatus as claimed in claim 1, wherein the light source further includes an optical arrangement and wherein each pulse of light from the light source is defocussed when it enters the target volume.
3. An apparatus as claimed in claim 2, wherein the defocussed light is a collimated sheet of light illuminating the target volume.
4. A method for capturing time-separated image samples of fluid flow in a target volume, the volume being seeded with a plurality of tracking particles, the method comprising: delivering a first pulse of light at a first time and a second pulse of light at a sccond time, each pulse of light having a predetermined peak frequency; capturing, in a plurality of colour camera units, a first set of component intensity samples corresponding to images of the tracking particles in the target volume while being illuminated by the first pulse of light and a second set of component intensity samples corresponding to images of the tracking particles in the target volume while being illuminated by the second pulse of light, the capture of the respective sets of component intensity samples being synchroniscd with the delivcry of the respective pulses of light capture, wherein the tracking particles are fluorescent type tracking particles having an absorption frequency band substantially including the predetennined peak frequency and the camera units each generate respective colour images, each colour image comprising a plurality of component intensity samples.
5. A method as claimed in claim 4, wherein each pulse of light is defocussed when it enters the target volume.
6. A method as claimed hi claim 5, wherein the defocussed light is a collimated sheet of light illuminating the target volume.
7. A method for determining velocity components of tracking particles in a target volume, the tracking particles being fluorescent type tracking particles having individual colour characteristics, the method comprising: generating a first set of component intensity samples at a first time; identifying discrete particles in each component intensity sample in the first set; determining a correspondence between a discrete particle identified in a first component intensity sample of the first set and a discrete particle identified in a second component intensity sample of the first set, said correspondence being dependent upon the colour characteristics of the discrete particles identified in the first set of component intensity samples, generating a second set of component intensity samples at a second time; identi'ing discrete particles in each component intensity sample iii the second set; determining a ftTrther correspondence between a discrete particle identified in a third component intensity sample of the second set and a discrete particle identified in a fourth component intensity sample of the second set, each further correspondence being dependent upon the colour characteristics of the discrete particles identified in the first set of component intensity samples, matching particles for which correspondence is determined in the first set to particles for which correspondence is determined in the second set, each match being dependent upon the colour characteristics of the correspondence particles; and determining the change in location of each of the matched particles for which a matching particle is found in the second set of samples, thereby determining the velocity components for the respective particles.
8. A method as claimed in claim 7, wherein the step of determining a correspondence comprises: pairing the discrete particle identified in a first component intensity sample of the first set and the discrete particle identified in a second component intensity sample of the first set when the particles have pixel locations within a predefined tolerance of one another; and validating the pairing by confirming that the paired particles have substantially the same colour characteristics.
9. A method as claimed in claim 7, wherein the step of determining a correspondence comprises: pairing thc discrctc particic idcntificd in a first cornponcnt intcnsity sampic of thc first sct and the discrete particle identified in a second component intensity sample of the first set when the particles have substantially the same colour characteristics: and validating the pairing by confirming that the paired particles have pixcl locations within a prcdcfincd tolcrancc of onc anothcr.
10. A method as claimed in any one of claims 7 to 9, wherein the step of matching particles comprises: associating particles for which correspondence is determined in the first set to particles for which correspondence is detennined in the second set when the correspondence particles have substantially the same colour characteristics.
11. A mcthod as claimcd in any onc of claims 7 to 9, whcrcin thc stcp of matching particlcs comprises: associating particles for which correspondence is determined in the first set to particles for which correspondence is determined in the second set when the correspondence particlcs havc a mefric of location relative to thcir ncar neighbours which is within a predetermined tolerance or each other; and validating association by confirming that the associated particles have substantially the same colour characteristics.
12. A mcthod as claimed in any onc of claims 7 to ii, furthcr comprising: generating a third set of component intensity samples at a third time; identiing discrete particles in each component intensity sample in the third set; determining a correspondence between a discrete particle identified in a fifth component intcnsity sample of thc third sct and a discrctc particle identified in a sixth componcnt intcnsity samplc of the third set, said correspondence being dependent upon the colour characteristics of the discrete particles identified in the third set of component intensity samples, matching particles for winch correspondence is detennincd in the second set to particles for which correspondence is determined in the third set, each match being dependent upon the colour characteristics of the correspondence particles; and determining the change in location of each of the matched particles for which a matching particlc is found in thc third sct of samples, thcrcby detcrmining thc velocity componcnts for the respective particles.
13. A method as claimed in any one of claims 7 to 12, wherein the step of determining the change in location of thc matched particles comprises: splitting the targct volume into a plurality of interrogation volumes, and perform a three-dimensional cross-correlation in the discrete particle locations to determine the motion of the interrogation volumes.
14. An apparatus for determining velocity components of tracking particles in a target volume, the tracking particles bcing fluorescent typc tracking particles having individual colour characteristics, the apparatus comprising: means for generating a first set of component intensity samples at a first time; means for identi'ing discrete particles iii each component intensity sample in the first set; means for determining a correspondence between a discrete particle identified in a first component intensity sample of the first set and a discrete particle identified in a second component intensity sample of the first set, said correspondence being dependent upon the colour characteristics of the discrete particles identified in the first set of component intensity samples, means for gcncrating a sccond sct of component intcnsity samples at a sccond time; means for identiring discrete particles in each component intensity sample in the second set; means for determining a further correspondence between a discrete particle identified in a third component intensity sample of the second set and a discrete particle identified in a fourth component intensity sample of the second set, each further correspondence being dependent upon the colour characteristics of the discrete particles identified in the first set of component intensity samples, means for matching particles for which correspondence is detemiined in the first set to particles for which correspondence is determined in the second set, each match being dependent upon the colour characteristics of the correspondence particles; and means for detennining the change in location of each of the matched particles for which a matching particle is found in the second set of samples, thereby determining the velocity components for the respective particles.
15. An apparatus as claimed in claim 14, wherein the means for determining a correspondence comprises: means for pairing the discrete particle identified in a fir st component intensity sample of the first set and the discrete particle identified in a second component intensity sample of the first set when the particles have pixel locations within a predefined tolerance of one another; and means for validating the pairing by confirming that the paired particles have substantially the same colour characteristics.
16. An apparatus as claimed in claim 14, wherein the means for determining a correspondence comprises: means for pairing the discrete particle identified iii a first component intensity sample of the first set and the discrete particle identified in a second component intensity sample of the first set when the particles have substantially the same colour characteristics; and means for validating the pairing by confirming that the paired particles have pixel locations S within a predefined tolerance of one another.
17. An apparatus as claimed in any one of claims 14 to 16, wherein the means for matching particles comprises: means for associating particles for which correspondence is detennined in the first set to particles for which correspondence is determined in the second set when the correspondence particles have substantially the same colour characteristics.
18. An apparatus as claimed in any one of claims 14 to 16, wherein the means for matching particles comprises: IS means for associating particles for which correspondence is determined in the first set to particles for which correspondence is determined in the second set when the correspondence particles have a metric of location relative to thcir near neighbours which is within a predetermined tolerance or each other; and means for validating association by confimiing that the associated particles have substantially the same colour characteristics.
19. An apparatus as claimed in any one of claims 14 to 18, further comprising: means for generating a third set of component intensity samples at a third time; means for identiing discrete particles in each component intensity sample in the third set; means for determining a correspondence between a discrete particle identified in a fifth component intensity sample of the third set and a discrete particle identified in a sixth component intensity sample of the third set, said correspondence being dependent upon the colour characteristics of the discrete particles identified in the third set of component intensity samples, means for matching particles for which correspondence is determined in the second set to particles for which correspondence is determined in the third set, each match being dependent upon the colour characteristics of the correspondence particles; and means for determining the change in location of each of the matched particles for which a matching particle is found in the third set of samples, thereby determining the velocity component for the respective particles.
20. An apparatus as claimed in any one of claims 14 to 19, wherein the means for determining the change in location of the matched particles comprises: means for splitting the target volume into a plurality of interrogation volumes, and means for perform a three-dimensional cross-correlation in the discrete particle locations to determine the motion of the interrogation volumes.
21. A method for determining velocity components of tracking particles in a target volume substantially as hereinbefore described with refcrence to the accompanying drawings.
22. An apparatus for determining velocity components of tracking particles in a target volume substantially as hereinbeforc described with rcfcrenec to the accompanying drawings.