In certain industries, such as the pharmaceutical and paint industries, it is important to be able to determine the range and distribution of sizes of particles in a sample. In this way, for example, a powdered sample may be characterised so that an end-user of the powder has accurate knowledge of the different sizes of particles to be found in the powder, as well as the distribution of such different sizes within the sample.

Previously, sizing techniques for solid particles having diameters greater than one micron have involved sedimentation, sieving, Coulter counting, direct imaging or laser diffraction. However, such methods have not been adapted to bias (and thus make more accurate) results against the fact that usually not all of the particles in a sample are detected and thus sized. Sedimentation, sieving, Coulter counting and laser diffraction techniques all involve the disadvantage that every particle is measured. Further techniques such as these that do not involve actual imaging suffer from the added disadvantage that the operator is not able to see the particles being sized.

In addition, Nishino et al (Measurement Science & Technology Vol. 11, No. 6, page 633) describe measuring both size and velocity of solid particles, but again no biasing of the size measurements was made according to the probability of actually detecting each particle.

Further, in the known technique of phase-doppler anemometry, the velocity of droplets in sprays (i. e. liquid phase in air) has been measured, but such measurements of droplets do not involve imaging perse.

It is therefore an object of the present invention to overcome the disadvantageous lack of accuracy of the prior art methods of sizing particles, to allow provision of more correct values of particle size distribution within a sample via novel imaging techniques.

Thus, in a first aspect, the present invention provides a method of measuring the distribution of sizes of particles within a moving sample, comprising: (a) illuminating a measuring region through which a sample of particles is passed; (b) capturing one or more first image (s) of individual particles within the sample ; (c) capturing one or more second image (s) of each individual particle for which a first image has been captured, and measuring the velocity of each such particle by comparing the first and second images; and (d) determining from either, or both of, the first and second images the size of each individual particle for which both said images are captured; whereby each velocity measurement is used to correlate the contribution of each sized particle to the total size distribution of all particles in the sample, whether sized or not, and so enable calculation of that total size distribution.

Typically, the second images are of some or all of the particles.

The term"particle"is herein defined to mean either (a) a solid unit, which may be dispersed within air or another gas, or within a liquid, or supported by a solid surface; or (b) a liquid droplet dispersed within another liquid having at least one differing optical characteristic.

The present invention makes use of the fact that particles that travel faster through a measuring region are less likely to be detected than those particles that travel more slowly through the region. Likewise, if particles are detected repeatedly, a faster particle will be detected fewer times. Thus, each detected incidence of a faster- moving particle should be given greater statistical weighting, than each corresponding incidence of a slower-moving particle, when calculating the distribution of particle sizes within a sample passing through the measuring region.

Each detected particle can be given a weighting (towards the overall size distribution of the sample) that is proportional to the reciprocal of the time that the particle spends in the measurement region. This time period can be calculated from the velocity component of the particle, and a knowledge of the size of the measurement region in the direction of the velocity component. Said size may in turn depend on the size of the particle and its position within the measurement region.

Large particles may be rejected if they touch the edge of the measuring region, whereas smaller particles may be accepted when not so touching, their centres lying in a larger area than the centres of larger particles.

The specific technique used for measurement of the velocity may be any one of several described below, but whichever method is chosen, more accurate distributions can be obtained by such measurement and by subsequently weighting the incidence of each detected particle accordingly.

In a second aspect, the present invention provides an imaging apparatus for measuring the distribution of sizes of particles within a moving sample, comprising: (a) one or more illuminating elements for illuminating a measuring region through which a sample of particles is passing; (b) one or more first capturing cameras for capturing one or more image (s) of individual particles within the sample ; (c) one or more second capturing cameras for capturing one or more second images of each individual particle within the sample for which a first image has been captured, and for measuring the velocity of each said particle by comparing the first and second images; (d) a sizing means for determining from either, or both of, said first and second images the size of each particle for which both said images are captured; and (e) a processor for statistically weighting the incidence of each sized particle in proportion to its measured velocity, adapted to calculate the total size distribution of all particles within the sample, whether sized or not.

Optionally, the one or more first and second capturing cameras may be the same camera (s).

When the sample is a powder of solid particles, the particles are typically (a) dispersed in air or another gas; (b) dispersed in a liquid ; or (c) resting upon, or attached to, a stationary or moving sheet of material (e. g. a conveyor belt).

When the sample consists of liquid particles, these particles will be dispersed in another liquid having at least one different optical property.

If the particles of the sample are solid and dispersed in air or another gas, the system preferably comprises a supply of flowing, pressurised gas and a mechanism for feeding the solid particles into the gas flow. For example, such a mechanism may be used to drop solid particles into the gas flow. Alternatively, the gas can be stationary and the solid particles may be fed into, and free-fall through, the gas. For example, solid particles may be arranged to fall off the end of a vibratory feeder into a stationary gas.

When the particles of the sample are dispersed within a liquid, the system preferably comprises a liquid flow guide, having optical access in the measuring region and a means for introducing the sample into the liquid. Such a means is necessary if the sample is not already in a liquid dispersion. An example of a sample already in a liquid dispersion is a clay or slurry.

The present invention will now be described in further detail with reference to the following examples as illustrated by the accompanying drawings, in which: Figure 1 shows the formation of an image with a line-scan camera.

Figure 2 illustrates velocity and aspect ratio measurement with a line-scan camera.

As mentioned above, the system comprises one or more illumination elements for illuminating the measurement region and examples of suitable elements include : continuous light sources, such as daylight from the sun, continuous- wave laser, or a continuous Light-Emitting Diode (LED) illumination ; quasi-continuous light sources, such as a tungsten filament lamp or a fluorescent electrical discharge tube; and pulsed light sources, such as a flash-lamp, pulsed laser, temporally- incoherent light source optically-pumped by a laser, or an LED.

In all the above examples, the light from the light source may be, from time to time, interrupted or diverted from reaching the measurement region by employing a blocking technique. Examples of means that may be used in the latter technique include : a mechanical chopper, a mechanical shutter, means for interrupting the electrical power supply to the light source, a rotating mirror, and an accousto-optic or electro-optic modulator.

The light output reaching the measurement region typically consists of a range of wavelengths from one or more of the visible, infra-red and ultra-violet spectral regions. Optionally, optical elements such as interference filters or absorption filters are used to select the wavelengths present.

The light may, although need not have, a defined polarisation state and, as desired, a means can be used to alter or select the polarisation state, such as a polaroid sheet, wave-plate, or a dielectric beam-splitter.

The system also comprises one or more sizing means for capturing one or more images of individual particles within the sample, and for determining from said images the size of each of some or all of the individual particles for which such images are captured.

Typically, this sizing means comprises a mechanism for projecting one or more real images of objects in the measurement region onto an electronic image sensor. Such mechanism can, in one embodiment, comprise an arrangement of one or more lenses, mirrors and pinholes. This mechanism can be placed in any one of a variety of places in the path of the light that is emitted from the light source and is directed towards the image sensor.

In one option, a means is present that can be used to alter the polarisation state of the light after it has interacted with the sample, but before it reaches the image sensor. Examples of such a means include: a polaroid sheet, a wave-plate, and a dielectric beam-splitter.

In another option, a means is present that can be used to alter the wavelength spectrum of the light after it has interacted with the sample, but before it reaches the image sensor. Examples of such a means for achieving such effects include: interference filters and absorption filters.

Preferably, an electronic image sensor is present, usually comprising eight or more individual light-sensing elements. Each element uses photons to modify the electrical properties of electrons contained in its structure, and thus provides an electronic system as a means for sensing the modified electrons and giving out a reading to a digital computer, whether the data is transmitted in analogue or digital form.

Examples of such electronic image sensors include CCD or CMOS cameras, whether line-scan, area-scan or progressive scan, and whether interlaced or not.

When a line-scan camera is used, a two-dimensional image is formed by taking one or more one-dimensional line-readouts obtained at different times and combining them, using time as the second dimension of the image (see Figure 1).

In a further embodiment, the option of applying one or more image processing techniques to the images is employed. For example, such processing may involve the subtraction of, or division by, a reference image; adjustment of contrast within the image; use of a look-up table for modifying individual pixel values ; or adjustment of the sensitivity characteristics of the system.

Typically, the system comprises a digital computer or digital processing unit, as a means for processing these images. Specific software is generally used that identifies individual particles on the image by comparison of their intensity on the image with the intensity of the background of the image, or with some other intensity that may be different in different parts of the image.

The processor used preferably measures a variety of features of the images of the individual particles detected, such as one or more of: Pixel area ; Perimeter ; Longest chord; Gradient of intensity at the edge of the particle ; Extremes of intensity found within the image of the particle ; Colour or relative wavelength intensities; and Variations of any of the parameters with regard to wavelength or polarisation of the light incident on or transmitted from the sample region.

Typically, the processor uses one or more of the preceding features to deduce the following characteristics of the particles being measured: Equivalent circular diameter (defined for example as smallest circle containing whole particle, largest circle contained by particle, or circle having same area as image of particle) ; Equivalent volume of spherical particle of that diameter; Length and radius of cylinder having the same pixel area and perimeter as the particle ; Dimensions of any other shape having the same pixel area and perimeter (or some other combination of measured features) as the particle ; The velocity of the particle (which use the direction dependent motion blur of the particles, or may require comparison with another image, or between two areas in the same image); and In the case of a system in which the images of particles are paired between or within images if they are estimated by the software to be images of the same particle, which other particle in this or another image is paired with a particular particle.

The processor assigns a statistical weighting to each particle and uses this weighting to combine the measured, or deduced, features of individual detected particles to produce an estimate of the distribution of sizes for all particles in the sample, whether detected or not.

In the simples case this weighting can be chosen to be unity for each particle (and thus need not be explicitly calculated by the processor). In some cases this weighting may be chosen to be zero, for example if a particle is touching the edge of the image, or if it is deemed out of focus. In the latter case the particle's contribution to the ensemble is not added.

In general, however, the weighting is chosen be neither zero nor unity and the choice is made depending on the following types of factors: The area of the particle (and hence its probability of touching the edge of the image); The size of the particle and so its probability of being accepted as in- focus; and The measured or deduced velocity of the particle (since fast moving particles may slip past the measurement region between exposures).

Whereas weighting based upon area and size of the particle are both techniques known in the prior art, weighting based upon velocity has not been previously described within the imaging field.

Several factors may chosen in combination when making the statistical weighting, and in such cases, the relevant components of the weighting are multiplied together to make the total weighting for that particle.

Measuring the velocity involves forming more than one image of a particle. In the simples form, two images are formed separated in time by a known and/or constant time. The velocity is proportional to the distance travelled by the particle divided by the time between the exposures.

An algorithm is required to pair the image of particle formed in the first exposure with that formed in the second exposure. This algorithm can take the form of particle tracking (usually requiring a knowledge of the predominant flow direction), or it can be a correlation based algorithm that detects the movements of patterns, or it can be a hybrid combination of both the former types of algorithms.

When such a two-image technique is used, two images of each particle that are separated in time are required. Examples of obtaining this effect include : A non-continuous light source, interrupted continuous and/or non- continuous light is used together with the sensitivity of an electronic camera. This allows each particle to be exposed at least twice and that both exposures are formed on the same image read out of the camera.

This solution is cost-effective, but has the disadvantage that unnecessary ambiguity in the resulting images can occur, since it is not possible to infer whether any given feature on the image was produced in the first or the second exposure.

Two or more independent images may be obtained by using a system in which there is no overlay of the images produced by the two exposures. Several types of camera may be used and various techniques for operating them may be employed to achieve this result.

These include : Use of a frame-transfer camera, for example that has been designed for use in particle-image velocimetry (PIV). In this type of camera, the entire image frame is transferred from light sensitive pixels to a set of non-light-sensitive pixels, in parallel. The first image is read out during the exposure of the second image. The time for which the camera is light sensitive in the first and second images need not be the same, and if there are more than two exposures, the intervals between the exposures may vary.

An interlaced exposure camera can be used, the output fields of which may be separated, with one field forming the first image, and the second field forming the second image.

A camera can be used in which some of the pixels are sensitive to particular wavelengths or polarisations (e. g. a colour camera) by coding the two exposures by polarisation or colour. In this case the camera may consist of more than one array of image sensing components (e. g. in the case of a 3-chip colour CCD camera).

In particular, it is desirable to take images in closely time-spaced pairs, but to wait a longer time to take the next pair of images. This method has the advantage that the same particle appears in a similar position in the two images obtained, and so is easy to pair each particle with its equivalent in the other image. Further, time is also allowed in this method for introducing a completely, or largely, new sample into the measurement region before the next exposure.

If the system speed is limited by the time taken to process images, or the rate of image capture, then such a method represents the most efficient way to time the images, if the number of particles is to be maximised.

Taking paired images in this way is known as"frame straddling" : the exposures falling at the end of the light-sensitive time for one camera frame, and at the beginning of the following one.

An alternative technique for measuring the velocity can be used in the case of a line-scan camera. In this case two lines are used, separated in space. By comparing the arrival times of the particle images at the two lines of sensors, the velocity can be deduced.

Arrival times can be measured in various ways including: time of arrival of the first sub-threshold pixel of a particle image, arrival time of the last such pixel, the average of these times, arrival time of line containing the most dark pixels for that particle, and arrival time of"centre-of-mass"of each particle's image.

Interpolation of intensities between pixels with the same and consecutive lines may be used to increase the effective resolution of the measure of the arrival time.

In the case of a line-scan camera, the velocity can also be used to estimate the correct aspect ratio (see figure 2) for the image formed since fast moving images appear"squashed"compared to slow-moving images (see Figure 1).

In Figure 1, a line sensor is shown with a particle moving across it with the motion of the real image being in direction A. Line-scan image B is of a fast moving particle, whereas line-scan image C is of a slow moving particle. Image D represents a reconstruction of a fast moving particle, whereas image E shows the reconstruction of a slow moving particle.

In Figure 2a, a camera with two line sensors is shown, with a fast (F) and slow (G) object approaching. Figures 2b and 2c respectively show images from line-scan sensors 1 and 2. Interval x is relatively long, since it represents the longer time between arrival times of the slow object (G) image. Interval y, by contrast, is relatively short, since it represents the shorter time between arrival times of the fast object (F) image.

Instead of separating the lines of sensors, it is also possible to give two closely spaced lines (or overlapping lines) selective sensitivity according to wavelength and/or polarisation. This allows the shifting of the images (rather than one of the sensors) according to wavelength and/or polarisation (for example, by the use of dichroic mirrors, or polarising beam-splitters, or both).

The system generally also comprises software and hardware for providing a means of displaying and storing the distributions obtained. Optionally, the system may comprise software and hardware for allowing the storage and subsequent display, re-analysis, or transmission of the images. For example, the system can allow for provision of overlaid or embossed information about the experiment including, but not limited to, the date and time, length or area scales, user-entered experimental details, individual particle information, and so forth.

The present invention thus provides a method and apparatus for accurately measuring the size distribution of particles within a sample, whereby the imaged particles are sized and given a weighting towards the overall sample distribution, thereby increasing the accuracy of calculating that distribution.