Method and apparatus for determining the flow rate of a viscous fluid stream

Abstract

A method of determining the flow rate of a viscous fluid stream comprises passing the fluid stream through a field of view of an electronic imaging device, which produces output signals representing images of the fluid stream at successive times t.sub.0, t.sub.1, . . . t.sub.n. Signals representing an image of the fluid stream at a first time t.sub.i are stored in a first memory, while signals representing an image of the fluid stream at a second time t.sub.j, are stored in a second memory. The time t.sub.j -t.sub.i is less than the time taken for a point on the fluid stream to cross the field of view of the imaging device. The stored signals are then analyzed to derive width signals W(y) representing the width of the fluid stream as a function of y, where y represents a coordinate in the direction of fluid flow, and the first and second stored signals are compared to identify a common point on the fluid stream in the respective images represented thereby. A signal representing the flow rate of the fluid stream is derived from the width signals W(y) and signals representing at least the y coordinates y.sub.1, y.sub.2 of the common point in the respective images. This method permits the continual monitoring of the volumetric flow rate while taking into account fluctuations in the width of the stream.

Description

BACKGROUND OF THE INVENTION

1. Field of the Invention

This invention relates to an apparatus for determining the volumetric flow rate of a viscous fluid stream, such as a molten glass stream.

2. Description of the Prior Art

In the manufacture of glass fiber, it is important to know accurately the volumetric flow rate, known as the pull rate, of the molten glass emerging from the heated forming bushing. The pull rate must be varied by continually adjusting the heat input to the forming bushing to maintain the characteristics desired in the finished fiber.

An apparatus is known for determining the flow rate of a molten glass stream by detecting bubbles naturally present in the stream and measuring the time taken for a bubble to flow between two predetermined points. Such apparatus is disclosed, for example, in U.S. Pat. No. 4,517,845 to Gullfiber. This apparatus employs a pair of spaced photodetectors which generate pulses as a bubble passes, and which after suitable shaping can be used to measure the time between the passage of the bubbles past the respective detectors. From the flow velocity thus measured, the volumetric flow rate can be determined by measuring the width of the fluid stream with the aid of a linear array of photodiodes.

While the apparatus described is capable of making a rough approximation of the volumetric flow rate, it suffers from inaccuracies due to the fact that in reality the width of the fluid stream varies as it flows between the photodetectors, leading to inaccuracies in the calculated volumetric rate.

Various other systems for measuring flow rate are known and are described, for example, in the following patents:

U.S. Pat. No. 4,877,436, I. Sheinkop, discloses an arrangement where an auxiliary falling stream of molten glass passes through a laser beam and the diameter of the stream is measured along with the time taken for air bubbles or `seeds` to pass through the beam. This patent depends on the detection of specular reflection from bubbles passing through a laser beam. As it requires lots of seeds, it is not suitable for highly refined glass.

U.S. Pat. No. 4,205,973, W. H. Ryan (Owens-Illinois Inc.), discloses an arrangement whereby the rate of fall, volume and section of a gob of glass are measured by two vertically spaced photodetectors and a pair of cameras arranged at 90 degrees in a horizontal plane. This arrangement suffers from the same disadvantage as Gullfiber mentioned above.

U.S. Pat. No. 3,739,636, (AA Versaci), discloses means for injecting a bubble into a fluid conduit, measuring the time taken to pass between two optical detectors and computing the flow rate. The arrangement suffers from the same disadvantage as Gullfiber.

A number of other arrangements for measuring fluid flow rate are disclosed in U.S. Pat. No. 4,866,639, R. J. Adrian (University of Illinois); U.S. Pat. No. 4,737,178, L. G. Relfschneider et al., (Owens-Corning Fiberglas Corp.); U.S. Pat. No. 4,837,708, T. B. Wright; U.S. Pat. No. 4,331,132, S. Mukasa (Olympus Optical Co., Ltd.); U.S. Pat. No. 4,554.832 - T. Hasegawa et al. (Nippon Furnace KK Ltd.); U.S. Pat. No. 4,925,296, J. Reichmuth; GB 2083612, P. H. Casswell; DE 1934583, C. T. Boe et al. (Instituttet fur Porduktudvickling Danmarks Tekniske Hojskole); JP56-147013A, H. Mukal (Klyouto Denshi KK); JP 55-040918A, K. Yamaguchi et al, (Niksuklsou KK).

An object of the present invention is to alleviate the disadvantages of the aforementioned prior art and provide a method and apparatus capable of more accurately providing the volumetric flow rate of a stream of viscous fluid, more particularly a stream of molten glass.

SUMMARY OF THE INVENTION

According to a first aspect of the present invention there is provided a method of determining the flow rate of a viscous fluid stream, comprising passing the fluid stream through a field of view of an electronic imaging means, said imaging means producing output signals representing images of said fluid stream at successive times t.sub.O, t.sub.l, . . . t.sub.n, storing in first memory means signals representing an image of said fluid stream at a first time t.sub.i, storing in second memory means signals representing an image of said fluid stream at a second time t.sub.j, where t.sub.j -t.sub.i is less than the time taken for a point on said fluid stream to cross the field of view of said imaging means, analyzing said stored signals to derive width signals W(y) representing the width of said fluid stream as a function of y, where y represents a coordinate in the direction of fluid flow, comparing said first and second stored signals to identify a common point on said fluid stream in the respective images represented thereby, deriving signals representing at least the y coordinates y.sub. l, y.sub.2 of said common point in the respective images, and deriving a signal representing the flow rate of said fluid stream from said width signals W(y) and said signals representing at least the y coordinates y.sub.1, y.sub.2 of said common point in the respective images.

In a preferred embodiment the output signals produced by the camera are digitized and stored at discrete addresses in the first memory means. The digitized signal at each address represents the intensity at a corresponding pixel in the image.

In order to determine the width of the fluid stream as a function of y, the images are first scanned line by line to identify the edges of the fluid stream. These are defined by changes in intensity exceeding a predetermined threshold relative to background intensity.

The stored images are then scanned to identify groups of adjacent pixels having values above a previously determined threshold. These groups are assumed to represent bubbles in the fluid stream. Once identified, the pixels of each group are set to a predetermined value, preferably 255, which designates them as parts of the bubbles.

The digitized signals are then compared to match the bubbles identified in the stored images. The bubbles are matched by determining the X and Y coordinates of all bubbles in the images. Since the lateral movement of bubbles in a fluid stream is very limited, bubble paths for which the difference in X coordinates exceeds a predetermined value, preferably 10 pixels, can be eliminated. Similarly, the distance travelled by the bubbles in the Y or longitudinal direction is limited by the limits on the changes in fluid stream velocity. For electronic scanning at 60 fields per second, the difference in the Y coordinates must lie in the range 320 and 380 pixels for a given lens magnification and image height of 512 pixels

From the fluid stream width and velocity, the mess flow rate can be determined from the following expression: ##EQU1## where, G.sub.dens =density of glass

T.sub.2 -T.sub.1 =Time interval between frames y.sup.1 =Y coordinate of bubble at time T.sub.1 y.sup.2 =Y coordinate of bubble at time T.sub.2 Width (y)=Stream width as a function of Y coordinate. K=Calibration constant.

The processing means is preferably in the form of a microprocessor with associated memory devices.

In another aspect, the invention provides an apparatus for determining the flow rate of a viscous fluid stream comprising an electronic imaging means having a field of view through which the fluid stream passes, said imaging means producing output signals representing images of said fluid stream at successive time intervals t.sub.o, t.sub.l, . . . t.sub.n, first memory means for storing signals representing an image of said fluid stream at a first time t.sub.i, second memory means for storing signals representing an image of said fluid stream at a second time t.sub.j, time t.sub.j -t.sub.i being less than the time taken for a point on said fluid stream to cross the field of view of said electronic imaging means, means for analyzing said stored signals to identify the edges of said fluid stream and derive signals W(y) representing the width thereof as a function of y, where y represents a coordinate in the direction of fluid flow, means for analyzing said stored signals to identify a common point on said fluid stream in said respective images and derive signals representing at least the y coordinates y.sub.1, y.sub.2 of said common point in the respective images, and processing means for deriving a signal representing the volumetric flow rate of said fluid stream from said width signals W(y) and said signals representing at least the y coordinates y.sub.1, y.sub.2 of said common point in the respective images.

The imaging means is preferably in the form of a CCD (charge-coupled device video camera). The camera is not restricted to operating in the visible region, and may, for example operate at infra-red or ultra-violet wavelengths.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will now be described in more detail, by way of example only, with reference to the accompanying drawings, in which:

FIG. 1 is a diagram illustrating the relationship between the camera and the molten glass stream;

FIG. 2 is a diagram diagrammatic view of a camera for use in a apparatus in accordance with the invention;

FIG. 3 is a diagram illustrating the movement of bubbles in the molten glass stream;

FIG. 4 is a block diagram in accordance with the invention; and

FIGS. 5a to 5j are flow charts illustrating the mode of operation of the processor in accordance with the invention.

DESCRIPTION OF PREFERRED EMBODIMENTS

In FIG. 1, a molten glass stream 1, which is to be drawn into a filament, flows out of a controlled forming bushing 2. The forming bushing is heated, and by varying the amount of energy applied, the viscosity of the glass flowing through it can be changed, thereby varying the characteristics of the glass flowing out of the bushing to determine the properties of the glass filament 1.

The glass stream 1 flows across the field of view of a video camera 3, which is conveniently a CCD (charge coupled device) camera with an electronic shutter. The CCD camera consists of an array of light sensitive cells that store a charge dependent on the intensity of light falling on them. The cells are sequentially accessed to provide an output signal that is a line by line scan of the image formed in the camera.

In accordance with standard video technology, the image is scanned at a rate of 30 frames per second, each frame being divided into two fields, the first field consisting of the even numbered lines and the second field consisting of the odd numbered lines, consequently 60 fields per second.

In a conventional video camera, the electronic shutter is programmed to operate once per frame. Each image is made up of the two interleaved fields. Consequently, although there are only 30 images per second, the actual flicker rate is 60 fields per second, and this creates the impression of continuous movement. In this apparatus, the camera's shutter is programmed to operate once per field, so that each interleaved field contains a different image.

The physical arrangement of the camera is shown in FIG. 2. The camera unit 3 contains the CCD camera 4 in an industrial, dust-tight camera housing with air cooling. The unit 3 has a transparent window 5 behind which is located lens system 6 consisting of a close-up lens 7, a 75 mm. zoom lens 8, and two times extender 9. This arrangement focuses the molten glass stream 1 onto the image plane of the CCD camera 4.

The purpose of the camera arrangement is to measure the width W(y) of the glass stream 1 (FIG. 3) and identify air bubbles 10 naturally occurring in the molten glass stream 1.

Referring now to FIG. 3, the volume of glass V contained between coordinates y.sub.1, y.sub.2, is given by the expression ##EQU2## where W(y) is the width of the glass stream as a function of y.

The mass flow rate F is therefore given by the expression ##EQU3## where T.sub.1, T.sub.2 are the times of measurement of the coordinates y.sup.1, y.sup.2, K is a calibration constant as defined above, and G is the density of the glass.

Consequently, since the incremental distance between each scanning line can be assumed to be a constant, the volumetric flow rate F can be derived from the sum of the widths of the glass stream 1 at each line between coordinates y.sup.1, y.sup.2, in accordance with formula (1) above, which is the discrete form of expression (3).

The apparatus for determining these parameters is shown in FIG. 4. In FIG. 4, there are two camera units 3, 3' that each operate independently on different glass streams. Each video camera unit 3, 3' generates sufficient information alone to enable the volumetric flow rate of each glass stream to be monitored. However, two or more streams, each with its respective camera unit 3, can be monitored simultaneously, and for this purpose the outputs of the camera units 3, 3' are fed to a video multiplexer unit 11, which time division multiplexes the outputs of the camera units 3, 3' into a common multiplexed output signal 12.

Each cell of the CCD array in the camera units 3, 3' represents a picture element, or pixel, of the image formed in the camera. The output signals of the CCD cameras consist of analog pulses having a value related to the intensity of light falling on the corresponding cell in the CCD array.

The multiplexed analog pulses output at 12 are fed to a video digitizer 13, which quantizes the analog pulses into 256 levels represented by 8 bit bytes (0 to 255). The output of the video digitizer 13 is fed directly to video display monitor 14, where each image can be viewed directly by the operator for monitoring purposes.

The output of the video digitizer 13 is also fed to a processing unit 15, which can conveniently be an IBM 80286 AT personal computer attached to a keyboard and monitor 16. The computer can be programmed in a manner to be described later, or alternatively the processing unit 15 can be a hard wired custom designed circuit implementing the processing functions to be described.

The processing unit 15 outputs a signal representing the calculated flow rate, which signal is passed to a digital-to-analog convertor 17 that outputs an analog signal representative of the volumetric flow rate and that in turn is connected to a bushing controller 18 that controls the energy applied to the bushing in accordance with the measured volumetric flow rate and a desired set point. The operation of the apparatus will now be described as follows:

The video signals from the cameras 3 are fed into the video multiplexer 11, which allows the video signal from any single camera to be selected by the computer 15. The output from the multiplexer 11 is fed into the high speed video digitizer 13, which captures single frames of video and presents them to the computer in digital form. The captured images can be viewed directly on the monitor.

The computer 15 analyses the video frames and calculates a volumetric flow rate. The resulting flow rate is displayed on the computer screen in the form of a number and a graph and is also sent to the digital-to-analog converter, which produces an analog output signal proportional to the calculated flow rate. In the case where more than one camera is used, the computer 15 selects the video signal from each camera in turn, performing the flow rate calculations and updating the analog outputs for each camera.

The digitizer 13 captures and digitizes a single frame from the camera. This results in an array of numbers with 512 elements vertically and 512 elements horizontally. Each element in the array represents the intensity of the image at the corresponding pixel in the image. The intensities range from 0 for minimum intensity to 255 for maximum intensity.

Each captured frame contains two separate images defined by the fields as described above. The first field is contained in the even numbered lines (0,2,4, . . . , 508,510). The second field is contained in the odd numbered lines (1,3,5, . . . , 509,511). These fields represent two images taken at an interval of one-sixtieth of a second.

The first stage in the analysis involves identifying which part of the image represents the glass stream. To locate the left edge of the stream, each line of the image is scanned from left to right until the intensity of the pixels exceeds a certain threshold value (see FIG. 5D). The right edge is similarly located by scanning from right to left. This operation is performed for every line in the image to generate a table representing the left and right locations of the stream edge for every point along its vertical axis. In the event that an edge is not detected, the left and right locations are set to zero.

The second step determines the average width of the stream by taking the average distance between the left and right edges of the stream for each point along its vertical axis (see FIG. 5E).

In order to enable the apparatus to adapt to changing lighting conditions, the threshold value is adjusted to lie midway between the average intensity of the stream and the average intensity of the background after each image is processed. The stream intensity is calculated as the average pixel intensity of the pixels lying within the boundaries of the stream as determined in step 1. The background intensity is calculated as the average intensity of all pixels outside of the stream boundary.

Next, the pixels representing bubbles within the stream must be identified. Bubbles show up in the image as bright regions within the stream. The only significant movement in the image is that of the bubbles. As a result, the bubbles can be highlighted by subtracting the pixels in the second field from the corresponding pixels in the first field. This resulting image represents the changes that have occurred between the two image fields. If the magnitude of the pixel difference is larger than a threshold value, then that pixel is assumed to be part of a bubble. If the difference is positive, the pixel belongs to a bubble located in the first field and if it is negative, it is located in the second fields. The pixels that are determined to represent bubbles in each of the fields are set to 255 and all other are set to zero (see FIG. 5F).

All of the bubbles, in both fields are then identified and their x, y locations and sizes determined. Bubbles with sizes above or below certain limits are ignored to eliminate noise.

Once the bubbles have been located, an attempt is made to identify bubbles that have shifted in location from one field to the next. This is accomplished by trying every possible combination of pairs of bubbles (one bubble from each field) and comparing their coordinates and sizes to see if they are the same bubble. Accurately matching the bubble pairs is possible because of the following:

1) Bubbles move very little horizontally between fields.

2) The distances the bubbles travel vertically between fields tend to lie within a small range.

3) The bubbles vary only slightly in size between fields. Using this information a bubble in the second field is considered to match a bubble in the first field the following conditions are met:

1) The difference between the X coordinates is less than 10.

2) The Y coordinate of the second bubble minus the Y coordinate of the first bubble lies between 320 and 380.

3) The ratio of the two bubbles sizes is no larger than 2 and no less than 0.5.

The actual number (320, 380 etc.) depends on the resolution of the camera and the magnification of the lens.

FIG. 3 shows an image of the stream with the two matching bubbles identified. Bubbles A and B have been rejected because of the above criteria. Bubble A is located too far to the right to be considered, while bubble B located directly below the first bubble is too small, and is rejected.

Once a pair of bubbles has been identified the flow rate can be calculated. Assuming that the glass stream is moving at the same rate as the bubbles, the amount of glass that has flowed past the camera during the time interval between fields is equal to the total volume of glass contained in the stream between the two Y coordinates of the bubbles. The flow rate can then be calculated from formula 1.

If no bubble pairs are found for this image, the flow rate can still be updated based on the width change alone using the equation: ##EQU4## where, Width=Average width of stream

LastFlow=Last flow rate calculated LastWidth=Average stream width at the time that LastFlow was calculated.

In this equation, the flow rate is estimated by multiplying the last determined flow rate by a factor that represents the ratio of the cross-sectional areas of the average width of the stream during the measurement to the area of the average width of the stream during the last measurement.

Once the digital signal representing the volumetric flow rate of the molten glass has been determined, this is output from the computer and converted back into an analog signal in the digital-to-analog converter 17. From there it is fed to PI (proportional integral) 18 controller, which in turn controls bushing controller 19 to set the energy input of the bushing 2. The control signal generated by the bushing controller 19 thereby controls the current flowing through the bushing 2, which in turn controls the glass flow rate.

The PI controller 18 adjusts the signal to the bushing controller 19 so as to maintain the glass flow rate at a given set point determined by the operator. The control is adjusted according to the formula:

Control=Control+P(E.sub.o -E.sub.l)+I (E.sub.o) where: Control=Control signal P=Proportional gain (constant) I=Integral gain (constant) E.sub.o =Present control error (Setpoint-Flow Rate)

E.sub.1 =Previous control error

In the event that the camera is unable to calculate the flow rate, the flow rate output from the camera is dropped to zero. When the control algorithm detects this, the PI algorithm is disabled and the control output is set to the average value of the control output over the previous ten minutes. When the output from the camera returns, the PI algorithm is re-enabled.

When the processing unit comprises an IBM AT2 86 computer, the following program can implemented:

______________________________________Initialize hardwareInitialize data structuresloopfor Each CameraCapture two successive images from selected cameraIsolate stream from background and determine W(y)Calculate average stream width (W)Isolate bubbles from streamDetermine x, y coordinates and sizes of all bubblesMatch bubble pairs from the two imagesif Bubble Pair FoundCalculate flow rate (F) by summing widths W(y)between bubble pairs using equation 1Replace F.sub.o with FReplace W.sub.o with WelseCalculate flow rate (F) using equation 2endifCalculate filtered flow (F.sub.f) to D/A converterUpdate user displayendforendloopend______________________________________

The detailed source code listing is shown in Appendix I.

The described method and apparatus allow the volumetric flow rate to be controlled on a continuous basis while taking into account variations in the width of the glass stream. In accordance with the set point determined by the operator, the output characteristics of the molten glass stream can be continuously maintained within desired tolerances. ##SPC1##

Claims

1. A method of determining the flow rate of a viscious fluid stream, comprising passing the fluid stream through a field of view of an electronic imaging means, said imaging means producing output signals representing images of said fluid stream at successive times t.sub.O, t.sub.l . . t.sub.n, storing in first memory means said signals representing an image of said fluid stream at a first time t.sub.i, storing in second memory means said signals representing an image of said fluid stream at a second time t.sub.j, the interval t.sub.j -t.sub.i being chosen so that said interval is less than the time taken for a point on said fluid stream to cross a field of view of said imaging means, analyzing said stored signals to derive width signals W(y) representing the width of said fluid stream as a function of y, where y represents a coordinate in the direction of fluid flow, comparing said first and second stored signals to identify a common point on said fluid stream in the respective images represented thereby, deriving signals representing at least the y coordinates y.sub.1, y.sub.2 of said common point in the respective images, and deriving a signal representing the flow rate of said fluid stream from said width signals W.sub.(y) and said signals representing at least the y coordinates y.sub.1, y.sub.2 of said common point in the respective images;

said output signals produced by said imaging means being digitized and stored at discrete addresses in said respective first and second memory means, the digitized signal stored at each address representing the intensity of the image at a corresponding pixel in the image;

and wherein the signals representing said stored images are first scanned line by line to identify the edges of the fluid stream defined by changes in intensity exceeding a predetermined threshold, thereby permitting the width of said fluid stream to be determined for each line;

said stored images being also scanned line by line to identify groups of adjacent pixels having values above a predetermined threshold, said groups being assumed to represent bubbles in the fluid stream, and the pixels of said identified groups are all set to predetermined values designating them as parts of bubbles.

2. A method as claimed in claim 1, wherein the signals representing said stored images are compared to match the bubbles identified therein, and the y coordinates of the matched bubbles between the images is determined from the pixel locations of a matched bubble in the first image and the second image.

3. A method as claimed in claim 2, wherein the bubbles are matched by determining the X and Y coordinates of all the bubbles in said images, comparing all the pairs of bubbles, and identifying bubbles as a matched pair if a) the difference in their X coordinates is less than a first predetermined value determined by the allowable lateral movement of a bubble in the fluid stream, b), the difference in their Y coordinates lies within predetermined limits defined by the minimum and maximum distance travelled by a point on the fluid stream at the minimum and maximum possible flow speeds, and c) the ratio of the two bubble sizes lies between predetermined limits.

4. A method as claimed in claim 3, wherein said first predetermined value is about 10, the difference in the predetermined limits of the Y coordinates lies between 320 and 380, pixels the ratio of the two bubble sizes is not less than about 0.5 and not more than about 2.0.

5. A method as claimed in claim 4, wherein the flow rate is calculated from the width of the fluid stream W(y) and the y coordinates y.sub.1, y.sub.2 of a matched bubble is determined in accordance with the expression: ##EQU5##

6. A method as claimed in claim 5, wherein when no matching bubble pairs are located in said images, a signal representing an updated flow rate is derived from the last determined flow rate and the square of the ratio of the average width of the fluid stream during the measurement and the average width of the fluid stream during the last measurement.

7. A method as claimed in claim 1, wherein said imaging means comprises a video camera.

8. A method as claimed in claim 7, wherein said imaging means comprises a CCD (charge coupled device) video camera.

9. A method as claimed in claim 1 which is used in the production of glass filaments wherein the signal representing the volumetric flow rate is used to control a forming bushing through which a molten glass stream is drawn.

10. An apparatus for determining the flow rate of a viscous fluid stream comprising an electronic imaging means having a field of view through which the fluid stream passes, said imaging means producing output signals representing images of said fluid stream at successive time intervals t.sub.O, t.sub.l, . . . t.sub.n, first memory means for storing said signals representing an image of said fluid stream at a first time t.sub.i, second memory means for storing said signals representing an image of said fluid stream at a second time t.sub.j, the intervals t.sub.j -t.sub.i being chosen so that said interval is less than the time taken for a point on said fluid stream to cross a field of view of said imaging means, means for analyzing said stored signals to identify the edges of said fluid stream and derive signals W(y) representing the width thereof as a function of y, where y represents a coordinate in the direction of fluid flow, means for analyzing said stored signals to identify a common point on said fluid stream in said respective images and derive signals representing at least the y coordinates y.sub.1, y.sub.2 of said common point in the respective images, processing means for deriving a signal representing the volumetric flow rate of said fluid stream from said width signals W(y) and said signals representing at least the y coordinates y.sub.1, y.sub.2 of said common point in the respective images;

and further comprising means for digitizing the output signals produced by said imaging means, said first and second memory means having a plurality of addresses at which said digitized signals are stored, each said stored digitized signal representing a corresponding pixel in the image;

wherein said means for analyzing said stored signals to identify the edges of said fluid stream scan the stored signals line by line to identify the edges of the fluid stream, said edges being defined by changes in intensity exceeding a predetermined threshold, and thereby determine the width of said fluid stream from the number of pixels between said edges;

and wherein said means for analyzing said stored signals to identify a common point on said fluid stream scan said stored signals to identify groups of adjacent pixels having values above a set threshold, said analyzing means thereby changing the values of the signals in the associated addresses to predetermined values, said predetermined values identifying the pixels represented thereby as parts of bubbles.

11. An apparatus as claimed in claim 10, comprising comparator means to compare the bubbles identified in said stored images to match up said bubbles, and the distance travelled by the matched bubbles between the images being determined from the number of pixels between the location of a matched bubble in the first image and its location in the second image.

12. An apparatus as claimed in claim 11 wherein said comparator means matches the bubbles by determining the X and Y coordinates of all the bubbles in said images, comparing all pairs of bubbles, and identifying bubbles as a matched pair if a) the difference in their X coordinates is less than a first predetermined value determined by the allowable lateral movement of a bubble in the fluid stream, b) the difference in their Y coordinates lies within predetermined limits defined by the minimum and maximum distance travelled by a point on the fluid stream and the minimum and maximum possible flow speeds, and c) the ratio of the two bubble sizes lies between predetermined limits.

13. An apparatus as claimed in claim 12 wherein said first predetermined value is about 10, the difference in the predetermined limits of the Y coordinates lies between 320 and 380 pixels, the ratio of the two bubble sizes is not less than about 0.5 and not more than about 2.0.

14. An apparatus as claimed in claim 13 wherein the flow rate is calculated from the width of the fluid stream and the distance travelled by a matched bubble in accordance with the expression: ##EQU6##

15. An apparatus as claimed in claim 14, further comprising means for deriving a signal representing an updated flow rate is derived from the last determined flow rate and the square of the ratio of the average width of the fluid stream during the measurement and the average width of the fluid stream during the last measurement when no matching bubble pairs are located in said images.

16. An apparatus as claimed in claim 10 which is used in the production of glass filaments, further comprising means for controlling a forming bushing through which a molten glass stream is drawn in response to the signal representing the volumetric flow rate and a signal representing a desired setpoint.

17. A method of determining the flow rate of a viscous fluid stream, comprising passing the fluid stream through a field of view of a video camera which is programmed to scan the image of the fluid stream in successive first and second fields, with the first field comprising even numbered scan lines and the second field comprising odd numbered scan lines, said camera having a shutter programmed to operate once per field, stored in first memory means signals from a first field representing images of the fluid stream at time t.sub.i and storing in second memory means signals from a second field immediately succeeding said first field and representing images of the fluid stream at time t.sub.j, the interval t.sub.i -t.sub.j being chosen so that said interval is less than the time taken for a point on the fluid stream to cross a field of view of said video camera; scanning said stored signals line to line to derive width signals W(y) representing the width of said fluid stream as a function of y for each line, where y represents a coordinate in the direction of fluid flow, comparing said first and second stored signals to identify a common point on said fluid stream in the respective images represented thereby, deriving signals representing at least the y coordinates y.sub.1, y.sub.2 of said common point in the respective images, and deriving a signal representing the flow rate of said fluid stream from said width signals W(y) and said signals representing at least the y coordinates y.sub.1, y.sub.2 of said common point in the respective images.