Download this paper - MS Word
| Lingying Zhao | Yuanhui Zhang | Xinlei Wang | G. L. Riskowski | L. L. Christianson |
| Ph.D., P.E. | Ph.D., P.E. | Ph.D., P.E. | ||
| Student Member | Member ASHRAE | Student Member | Member ASHRAE | Member ASHRAE |
| ASHRAE | ASHRAE |
Lingying Zhao is graduate research assistant; Yuanhui Zhang is associate professor; Xinlei Wang is graduate research assistant; and Leslie L. Christianson and Gerald L. Riskowski are professors in the Bio- environmental Engineering Research Laboratory, University of Illinois at Urbana-Champaign.
A measurement technique based on Particle Image Velocimetry (PIV) to measure quantitatively airflow patterns and distribution in ventilated airspaces is presented. Air laden with helium filled bubbles was illuminated by a two dimensional light sheet in a full-scale ventilated room. Images of bubbles visualized in the light sheet were recorded using a photographic camera. Relatively long camera exposure time makes it possible to record the bubble path in the designed time period. Image shift techniques remove the directional ambiguity. The photographic images were scanned into a computer and the digitized images were processed automatically using an image-processing program to extract flow field velocity information. The configuration, working principal, sample results, accuracy, capability and limitations of the technique are discussed in this paper. This measurement method is part of a larger study of aerosol spatial distribution, ventilation effectiveness and aerial contaminant control strategies.
Particle image velocimetry, airflow patterns, image processing, air quality, ventilation efficiency.
In the United States, people spend 93% of their lifetime and many animals spend their entire life indoors. Indoor air quality (IAQ) is increasingly recognized as important to occupants' health, well-being and productivity. Air velocity, flow patterns, and airborne contaminant distributions are among the most important factors affecting IAQ.
Airflow patterns affect the air-borne contaminant spatial distribution and comfort of building occupants in a ventilated air space. Improper indoor airflow patterns are frequently described as air drafts, insufficient ventilation, poor distribution, stuffiness, etc. Therefore, it is essential to study room air distribution of commercial and residential rooms, clean rooms, electronic and computer rooms, hospital rooms, greenhouses, and animal buildings (Christianson 1989). Because the air distribution and ventilation system determines the airflow patterns and thus affects the airborne contaminant level, understanding room air distribution is critical for design of ventilation systems and equipment.
Room airflow is complex and has multi-flow features including laminar boundary layers, highly turbulent diffuser jets, and low turbulent flow in the occupant region. Methods of studying indoor airflow include numerical simulation and experimental measurements. Many numerical models have been developed for room airflow study. Numerical simulation results vary greatly among models because of the simplified assumptions, insufficient validation by experiment data, and limited understanding of boundary layer conditions. Prototype experiment studies are very expensive and time consuming largely due to limitations of the current available measurement technologies and instrumentation.
A basic and long-standing problem in indoor air quality and environmental research is the lack of proper measurement techniques and instrumentation to describe quantitatively full-scale airflow in rooms. Difficulties include measurement of low air velocities, the direction of the velocities, high turbulent intensity, for large open spaces and complicated geometry in buildings. For low speed indoor airflow velocity measurements, the buoyancy effect makes it difficult to use thermal based sensors. The thermal anemometers commercially available are designed for air velocities higher than 30 fpm (0.15 m/s), which is above the indoor air velocities in many occupied zones. The disturbance to the airflow field created by the physical obstruction of the instrumentation and the sensors themselves is difficult to evaluate. Most previous researchers used hot wire anemometers to measure the velocity distribution in full-scale rooms. Laser Doppler Velocimetry (LDV) can measure low velocity magnitude and direction accurately without disturbance to the flow fields, but it can only measure one point at one time, and is expensive. For transient flows, point measurement results are difficult to interpret since the various spatial locations are sampled at different times and different flow conditions. Some researcher used LDV to measure velocity distribution in a reduced scale model room. For full-scale room measurements, LDV is difficult to set up. These gaps result in the lack of good data to validate computer models for indoor air flow and aerosol spatial distribution.
Therefore, non-intrusive, full scale, accurate and fast measurement techniques for low speed airflow in rooms are needed. A technique that uses particles and their images to measure flow velocities is called Particle image velocimetry (PIV). PIV is a promising technology to meet the needs of room air studies. PIV does not disturb the flow field, can measure the flow in full scale accurately, and has no low speed limitation. Two-dimensional PIV has been successfully developed in small-scale mechanical fluid studies (Adrian 1991). This technology needs to be developed for full-scale room airflow research.
The invention and development of PIV is originally for experimental fluid mechanics study (Adrian 1991). PIV measures a 2-D velocity vector map of a flow field at an instant of time by acquiring and processing images of particles seeded into the flow field. It is based on the principle that speed is equal to the displacement divided by the time interval. Particle speed is the representation for the flow field velocity.
Particle streak photography plus manual analysis of the image gave the semi-quantitative measurement results of flow field (Fage and Townend 1932). Modern developments in digital image processing and optical instruments make the image acquisition, process, and analysis more automatic and quantitative. Particle tracking velocimetry (PTV) or Particle streak velocimetry (PSV) are the extension developments of flow visualization. In PTV or PSV, the concentration of seeding particles should be dilute enough to form individual particle streaks, which can be analyzed automatically. In this operating mode, the spatial resolution of the method is not good enough to investigate complicated fine flow structures. Low-image-density PIV, high-image-density PIV, and Laser speckle velocimetry (LSV) are inventions to investigate complicated flow structures. In LSV, dense particles are seeded into flow and illuminated by a laser light sheet to form laser speckle images. By analyzing the double exposure laser speckle image using Yang's fringe method of interrogation (Burch and Tokarski, 1969; Stetson 1975), the speckle displacement information is extracted and then the velocity data are obtained. High-image- density PIV was established as an improvement mode of LSV to overcome the high-density particle seeding difficulty in practical situations by Pickering and Halliwell (1984) and Adrian (1984). With PIV, images of a group of particles or one particle falling into interrogation spots are analyzed by auto- correlation or cross correlation analysis methods depending on the image acquisition modes: one frame two exposures or two frame one exposure. Similarly, particle displacements are extracted to form velocity data information.
In practical application situations, according to the measurement requirements, the appropriate operating mode needs to be selected and specific system configurations are needed. In general, a PIV system consists of illumination, image acquisition, particle seeding, and image processing and data analysis sub-systems. Laser light is commonly used as the illumination source in PIV systems. Ideal tracer particles should be very small and follow the flow field. Different flow fields need different seeding particles. The design of the seeding particles was discussed by Schmitt et al. (1995). According to spatial resolution requirements, either photographic or CCD video cameras can be used to record the particle images. If the resolution is high enough, CCD video cameras are preferred. To resolve velocity direction ambiguity during the image acquisition process, an image shifting technique is needed (Adrian 1986b). Image interrogation techniques include auto-correlation and cross-correlation image processing methods. Commercial software and hardware products for PIV image interrogation are available. PIV started as a two dimensional velocity measurement method and is being developed to a three dimensional velocity measurement method. Stereoscopic photograph techniques have been employed to detect the three components of a velocity vector. Recently, holographic techniques have been researched to capture three-dimension information of a flow field.
Even though many PIV systems have been developed and some are becoming commercial products, most PIV experiments have been conducted on flows at very small scales, typically around 100 x 100 mm field of view, and fairly low turbulent flow (Adrian 1990). Enlarging the study scale is limited by the camera resolution and flow field illumination.
Brodie et al. (1993) used a CCD video camera (512x480 pixels), a project light sheet, and latex particles (200 to 400 ?m) to study low speed airflow. Velocity vectors were calculated based on the position of identical particles between the successive images. The resolution was not high enough to form clear images and the particle pairing was difficult to analyze because of seeding particle concentration variation. Akikazu et al. (1996) presented a PIV system for airflow measurement using smoke particles. A fast flow pattern-tracking algorithm and a qualitative flow field velocity distribution was described. Rasmus (1996), also used smoke particles to visualize flow fields in livestock buildings. Flow patterns were tracked to estimate the local velocity of the flow field. Some qualitative results were presented.
Muller and Renz (1996) presented a very expensive PIV system consisting of a 23 W Argon-Ion laser, a 25 m long light fiber, a 1012?1524 pixels resolution digital camera, a magnetic shutter synchronizer, and a helium bubble generator. Two-dimensional velocity measurements were conducted in a gymnasium with the dimensions of 15?3 m. An 1?1.5 m area was measured at one time. An interpolated airflow pattern was presented. However, further discussions about accuracy of the measurement were not given.
Scholzen and Moser (1996) developed a three-dimensional particle streak velocimetry system. In their study, three cameras, a 120 mm thick white light sheet, and a digital image processing program were employed to acquire the particle streak image and extract particle displacement information. One camera with a relatively short exposure time setting was used to recognize the particle streak direction. The other two cameras, which were in the same setting but put in different locations, formed stereoscopic photographs to obtain the three components of the velocity vectors. The method was tested in a 2.4 ? 1.7 ?1.2 m ventilated model space. Effective image area was 1.0 ? 0.8 m. Even though detailed image processing information was not discussed, one measurement with three images must result in large size data files and long data processing time. Good represented results were presented and showed that the method is promising for indoor airflow study.
A simple, inexpensive two-dimension room airflow PIV system has been developed (Figure 1). It consists of a test room, an illumination system with two sets of halogen lights with designed reflectors and cylindrical lenses, a flow seeding system with three helium filled bubble generators and designed bubble path tubes, an image acquisition system with a 4x5 photographic camera, an image shift system with a step servo-motor, and an image processing and interpolating system with a laser scanner, micro- computer, image processing software, and data analysis programs.
A room ventilation simulator (RVS) was used to simulate the ambient environmental conditions of the study. The RVS (Figure 2) consists of an outer room (9.1 x 12.2 x 3.6m ), which can simulate weather conditions from –25°C to 40°C any time during the year, and adjustable inner rooms to simulate ventilated buildings (Wu et al. 1989). The inner test room is of a modular design, so that different room configurations and sizes (up to 10x7x3 m) can be modeled conveniently. A 5.5 x 3.7 x 2.4 m test room has been set up within the RVS for this study. One long wall of the test room is made of glass to permit convenient optical access. The two short-walls contain four glass slits to transmit the illumination light. The other long wall, floor, and ceiling surfaces are painted with non-reflective black paint to form a good optical background. The top view of the test room location and configurations are shown in Figure 3.
The function of the illumination system is to visualize particles seeded into the airflow, so that images of the particles can be recorded and processed. In this study, white lights are used. The intensity of white light is inversely proportional to distance that the light travels. Twelve halogen lamps, each consisting of an aluminum chassis with a 1500-watt light bulb, a special designed reflector, and a cylindrical plano-convex lenses were mounted in two opposite walls of the test room to approximate uniform illumination. The reflector was partially painted into black. Therefore, it reflects the backward portion of light forward and absorbs the portion of lights that can not reach the lens in a designed angle. The cylindrical lenses installed in front of the light bulbs convert the twelve line light sources to a light sheet, which is 65 mm thick. To reduce heat accumulation and light scattering noise from the space around the lens, the two series of lights are put in two black, ventilated boxes.
Neutrally buoyant, helium filled bubbles, which are about 1 mm in diameter and are generated by three bubble generators. Bubbles are seeded through the room air inlet, ceiling, and one short wall through plastic tubes into the test room. Since the bubbles are neutrally buoyant, they follow the airflow. Therefore, bubble movement represents airflow and bubble velocities are used to represent the airflow velocities of the flow field, in which the bubbles travel.
To capture high-resolution images, a 4 x 5 photographic camera is used to capture the flow pattern. Light intensity, shutter speed, and lens aperture diameter control film exposure. According to the indoor airflow velocity range, the best exposure time under the described experimental conditions was found to be 1/4 s. Shutter speed is a measure of the distance particles travel when the shutter is open and cause an elongated image. The camera aperture was set at 8 based on the light intensity and the shutter speed setting.
When a picture of particles is taken on one frame with long exposure time, each particle will have a streak image on film. Since room airflow is most turbulent flows, a particle steak could be in any direction. It is difficult for the computer to recognize which end of the particle streak image is the start point and which end is the end point. To identify the direction ambiguity of room airflow, image shift techniques are used. During the image acquisition procedure, a precisely controlled step motor moves the film holder of the camera at a constant velocity. Therefore, a shift velocity is added to the camera and a still point object will form a streak image on the film. Velocity of the shift (Us) is determined by camera magnification (M) and maximum reverse flow velocity in the test room, and by analysis of sample test images. Because the maximum room air velocity is around 0.2 m/s, the camera is shifted by a step motor which moves at a constant speed of 0.3 m/s. The step motor is controlled precisely by a computer.
The photographic images are digitized using a laser scanner with resolution of 1016 pixels per inch. During the digitization, threshold value and contrast need to be selected carefully to ensure that the digitized image precisely reflects the original.
The digitized images are processed automatically using image-processing programs developed within commercial image processing software. The processed images were measured to extract raw particle position, angle and length information. Visual Basic programs were used to manipulate the raw data. Interpolated velocity data were smoothed by a software developed the continuity of the flow field. Graphic software was used to plot the velocity vector map and visually display the flow field.
The flow field is divided into sub-areas according to the flow field velocity range and flow characteristics. Sub-areas are divided by white threads and marked with symbols representing flow field locations and settings. Rulers and small LED lights are used to calibrate the camera, lens, and image shift system. Magnification ratio is selected against the ruler image. Coordinates of LED light images on film are corresponding to the spatial locations of the lights in the test room. Shifted images of the LED lights exactly show how much the shift velocity is. After the room ventilation system ran 30 minutes and the room airflow achieved stable state, bubbles were injected into the room through the particle seeding system. Lights were opened to illuminate only the bubbles within the light sheet. Image of particles visualized in the light sheet are recorded as streaks using the photographic camera moving at constant shift speed.
Key parameters to obtain quality images include seeding particle concentration (N), image acquisition magnification (M), f number of the lens (f#), camera exposure time (Δt), light sheet width, light intensity, and their combination.
The particle concentration, N, which is the average number of particles per unit surface, determines the average distance between particles δ (Agui and Jimenez, 1987) by:
(1)
In this study, the particle streak length is designated as Δ = 50 mm according to mean air velocities, camera exposure time, and spatial resolution requirements. Usually the mean distance between particles should be larger than the particle streak length to avoid an excessive number of intersections of particle streaks. Agui and Jimeze suggested δ = 1.5 Δ. The appropriate particle concentration can be obtained with equation (1).
The diameter of a particle image is determined by the diameter of the particle, the magnification, and the point diffusion of the lens. The point diffusion of the lens is determined (Adrian, 1991) by:
(2)
where
M = the magnification of the image acquisition system
λ = the wavelength of the light.
Then the diameter of the particle image, de, is calculated by:
(3)
where
dp = the diameter of the particles.
In this study, the point diffusion of the lens (ds) is 0.0145 mm, and the diameter of the particle image is 0.068 mm considering the point diffusion of the lens. Because the magnification of the image acquisition system is 1/15 and the particle diameter in our experiment is about 1 millimeter, the direct calculation from M x dp is 0.067mm. Since the two calculated values of dp have no significant difference, the lens diffusion effect can be ignored.
The following equation gives the quantitative evaluation of the camera focus to the illuminated particles (Adrian, 1991).
(4)
where
δz = the depth of field of the lens
Outside of the depth of field of the lens, the image is blurred by an amount exceeding 20% of the in-focus diameter (Adrian, 1991). In this study, the depth of field is 45.8 mm calculated from the above equation by choosing M as15, f# as 8, and λ as 0.7 μm. The light sheet is 65 mm. So, most particles in the light sheet can form clear images.
The bias caused by the thickness of light sheet and the three dimensional movement of particles are calculated by:
(5)
ΔX = the displacement of a particle image Δx = the displacement of a particle M = the magnification of camera lens X = the spatial location of particle in light sheet d0 = the distance between camera lens to the particle Δz = the out-of-plane image displacement of a particle
Particle streaks that are too long will reduce the spatial resolution and the detailed flow structures, but streaks that are too short will decrease the accuracy of the velocity measurement. According to the streak length and the flow field velocity, the camera exposure time can be determined. For the room airflow, the flow field is divided into a diffuser jet region and an occupant region according to mean velocity range. Because the typical velocity is from 0 to 0.2 m/s in the occupant region and from 0.6 to 5 m/s in the diffuser region, the exposure time is set to 0.25 s and 0.01 s, respectively. The dimension of the field of view is at least 1.5 x 2.0 m.
During the image acquisition process, the film is shifted by a servo-step motor to add a shift amount to the particle displacement to remove the particle streak direction ambiguity. In a 2-D situation, the velocity of a moving object is calculated by:
(6)
Δx = the displacement of a particle in the x direction Δy = the displacement of a particle in the y direction Δxs = the shift of images by moving the camera
Δx is possibly larger or smaller than zero depending on which direction the particle is moving. If a shift Δxs is added to the Δx to ensure that the sum result is larger than zero, the streak start point and end point can be easily recognized by the x coordinators of the points. After the particle streak direction is determined, the ?xs can be subtracted from the measured displacement in the x direction. Thus, actual velocity magnitude and direction are obtained. The shift (Δxs) can be determined by the maximum reverse flow velocity multiplied by the camera exposure time, Δt, and magnification, M. The resolution of 1016 dots per inch results in 40 pixels per millimeter on a digitized image and 3 pixels per millimeter on the actual flow field. The non-uniform background and image noises are removed using a high pass filter and a median filter. The streaks are enhanced by an edge enhancement function such as Soble operators. Morphological operation such as a dilation operator is used when necessary. Based on the gray value histogram analysis of particle streaks, the optimal thresholding is used to binary the image. Once the image is properly segmented, the center point, length, and the angle of the streak can be extracted automatically. Statistical percentile analysis is used to remove intersecting streaks and image noises. The data are calculated to obtain the velocities of each random location in the flow field.
Randomly located velocity data are interpolated to form uniform grid data. The grid size is determined by the measurement spatial resolution requirement. There is a trade off between the grid size and the measurement accuracy. A large grid size has large interpolation error and relatively small error of streak length measurement. A small grid size results in large streak length measurement error and small interpolation error. In summary, the spatial grid size is determined by the practical flow pattern, spatial resolution, and particle streak length. In this study, the grid size of 50 mm is selected based on the particle streak length and an acceptable spatial resolution to indoor airflow.
In interpolated velocity data, error vectors, which have a finite probability due to the image process, can be removed or replaced according to the neighborhood analysis. The vector modification can be based on an allowable range for the total velocity vector, an allowable tolerances for root mean square fluctuations, signal to noise ratio threshold, allowable tolerances in neighborhood median magnitude, and the alternative choices corresponding to other streaks in the grid.
The data analysis and measurement procedures outlined by Bendat and Piersol (1986) have been adopted in this project. The basic analysis for systematic bias and the variance is conducted for each step of the data recording and transmission procedures. The airflow pattern and velocity vector maps are then plotted according to the raw and processed velocity data.
The method has been tested by conducting room air distribution studies in a full-scale room simulating half of a full-scale swine building. Neutrally buoyant helium bubbles were used to track the indoor airflow. The densities of the bubbles and the air are the same. Continuous slot air inlet and outlet room ventilation settings with an air exchange rates of 9.5, 15, and 19.5 air changes per hour (ACH) were used to verify the feasibility of the method. The typical measurement results are velocity vector maps, an airflow pattern map, and air velocity distribution data.
The PIV method was operated in two modes: direct image acquisition and image acquisition with shift. Because in the some flow fields, the airflow has a dominant direction, direct image acquisition method is used to record the particle trajectories. With help of flow visualization, a clear flow velocity vector map can be obtained. Figure 4 (a) is an example of the raw image of a flow field with dominant flow direction. Figure 4(b) is the velocity vector map extracted directly from the raw image data (Figure 4a). Figure 5 shows the interpolated and smoothed air velocity vector map.
It is difficult for traditional measurement techniques to measure low air velocities because of thermal effects. The PIV measurement method can recognize very small velocities. Figures 6 to 8 show the ability of the PIV method to resolve low air velocities. Figure 6 shows bubble images in an occupied zone with an air exchange rate of 19.6 ACH. The magnification is 15 and the air velocity range is from 0.17 m/s to 0.197 m/s. Figure 7 shows bubble images in the same occupied zone with an air exchange of 15 ACH. The velocity range is from 0.02m/s to 0.1 m/s. Figure 8 shows image of bubbles in the same occupied zone with an air exchange of 9.6 ACH and velocities ranging from 0 m/s to 0.01 m/s.
In most occupied zones of the room, airflow is turbulent and the velocity range is 0 to 0.2 m /s as shown in Figures 6 to 8. Since the airflow may flow in any direction, image acquisition with shift mode must be used to clarify the direction ambiguity. Figure 9 shows a shifted image of the bubbles at a shift speed of 0.2 m/s. The shifted image has clear steak ends and can be measured accurately. The maximum shift ability is 0.4 m/s and covers the low velocity range in room airflow studies.
Some streaks in shifted image is light and can not be recognized accurately by the computer. Manual enhancement is made to several light streaks. The enhanced image is measured to extract the direction and length information. After the shift amount was subtracted, the original bubble movement information was recovered. Figure 10 shows a velocity vector map extracted from the shifted image. For turbulent flows, directly extracted images can not show clean flow patterns due to flow field seeding difficulty. Based on the continuity theory of the flow field, velocity vectors were interpolated. The interpolated velocity vector map shows clear flow patterns (Figure 11).
The intrinsic error sources for the whole measurement process are image acquisition, image processing, and velocity interpolation. Image acquisition errors include image digitization error, exposure time interval error, and tracking error, which is introduced because the seeding particles do not exactly follow the motion of the flow. Image processing errors refer to all the possible errors caused by image processing. Interpolation error comes from the attempt to translate the velocities of randomly known points to artificial grid points of the flow.
Because the film used to record the particle image has an approximate resolution of 300 lines/mm and the magnification is 15, one millimeter in the field of view is resolved by 45000 pixels. The errors caused in film recording is negligible. The particle image on film is digitized by a laser scanner using 1016 pixels/inch resolution. On this resolution setting, one millimeter in the field of view is resolved by 2.7 pixels and the uncertainty introduced by this resolution is approximately 0.4 mm. Since the average particle streak length is 50 mm, the error is approximately 0.8% of the measurement.
When the camera shutter is controlled electronically, the exposure time is precisely the same as the setting value. Mechanically controlled camera shutter systems usually have a maximum of 10% uncertainty. This will result in a maximum of 10% particle velocity measurement error.
The tracking error is analyzed by Agui and Jimeze (1987) and the error calculation is given as:
(7)
where
Stk = the Stokes number of the flow
Δ p = is the density of the particles
Δf = is the density of the fluid
In this study, the densities of the bubbles and the air are the same. So this error also is negligible.
During the image processing procedures, particle streaks are enhanced, noise is removed using specific filters, and the center of two end points of the streak are estimated. All these procedures result in a maximum of 4 pixels of uncertainty, which represents to 2 mm in the flow field. This the error caused by image processing is a maximum of 2.6%.
The interpolation error is determined by the grid size, the fluctuation of the velocity in the grid, and the velocity spatial gradient. The maximum possible error is the maximum fluctuation within the grid or the velocity gradient multiplied by grid size. According to the studies of Zhang (1991) and Riskowski (1993), this error can be estimated within 1%.
The exact accuracy of the PIV method will be determined by calibration. During the calibration process, a known flow field, in which airflow velocities are precisely controlled by a wind tunnel chamber, will be measured using the PIV airflow system. The overall error of the study is conservatively estimated within 15%. The mechanical shutter system of the camera is the largest error contributor. With an electronically controlled camera shutter system, the total error of the method can be reduced to 5%.
This work is supported by the Illinois Council for Food and Agricultural Research.
Adrian, R. J. 1984. Scattering particle characteristics and their effect on pulsed laser measurements of fluid flow: speckle velocimetry vs. particle image velocimetry. Appl. Opt. 23, pp. 1690-91?
Adrian, R. J. 1986. Image shifting technique to resolve directional ambiguity in double-pulsed velocimetry. Appl. Opt. 25(21), pp. 3855-58.
Adrian, R. J. 1990. Applications of particle image velocimetry.
Adrian, R. J. 1991. Particle imaging techniques for experimental fluid mechanics. Annual Review of Fluid Mechanics. 23, 261-304.
Agui, J.C. and J. Jimenez. 1986. On the performance of particle tracking. J. Fluid Mech. Vol.185, pp. 447-468.
Akikazu K.; Y. Katsuhito; I. Yoshio; K. Akira; K. Sinichi, and Y. Taro. 1996. PIV system for airflow measurement using smoke particles. Proceedings of 5th International Conference on Air Distribution in Rooms, ROOMVENT '96. PP. 421-426.
Bendat, J. S. and A. G. Piersol. 1986. Random data analysis and measurement procedures. John Wiley & Sons, Inc. ISBN 0-471-04000-2.
Brodie, J. R.; J. F. Reid; and T. Funk. 1993. A machine vision system for low velocity air pattern visualization. ASAE paper No. 933607.
Burch, J. M. and J. M. J. Tokarski. 1969. Production of multiple beam fringes from photographic scatters. Opt. Acta 15, pp.101-11
Christianson, L. L., ed. 1989. Building systems: Room air and air contaminant distribution. Atlanta: American Society of Heating, Refrigerating, and Air-Conditioning Engineers, Inc. Fage, S. and H. C. H. Townend. 1932. Turbulent motion near a wall. Proc. R. Soc. London Ser. A 135, pp. 656-77.
Muller, D. and U. Renz. 1996. Determination of all airflow velocity components by a particle image velocimetry system (PIV). Proceedings of 5th International Conference on Air Distribution in Rooms, ROOMVENT '96. pp. 413-419.
Pickering, C. J. D. and N. A. Halliwell. 1984. Speckle photography in fluid flows: signal recovery with two-step processing. Appl. Opt. 23, pp. 1128-29.
Rasmus L. 1996. Estimation of airflow in livestock buildings using image analysis. Proceedings of 5th International Conference on Air Distribution in Rooms, ROOMVENT '96. pp. 49- 64.
Riskowski, G. L.; J. B. Priest; S. E. Ford; and L. L. Christianson. 1993. Environment in animal regions. Livestock Environment IV, Fourth International Symposium, University of Warwick, Coventry, England. pp. 411-418.
Scholzen F. and A. Moser. 1996. Three-dimensional particle streak velocimetry for room Airflow with automatic stereo-photogrammetric image processing. Proceedings of 5TH International Conference on Air Distribution in Rooms, ROOMVENT '96. pp.555-562.
Schmitt T.; J.N. Koster; and H. Hamacher. 1995. Particle deslgn for displacement tracking velocimetry. Meas. Sci. Technol. 6, pp. 682-689.
Stetson, K. A. 1975. A review of speckle photography and inerferometry. Opt. Eng. 14, pp. 482-89
Wu, G. J.; L. L. Christianson; J. S. Zhang; and G. L. Riskowski. 1990. Adjustable room ventilation simulator for room air and air distribution modeling. Indoor Air '90, Proc. 5th international conference on indoor air quality and climate, 4, PP.237-242.
Zhang, J.S. 1991. A fundamental study of two dimensional room ventilation flows under isothermal and non-isothermal conditions. Ph.D. thesis, University of Illinois at Urbana-Champaign.
Figure 1. Schematic of the room air PIV system
Figure 2. Room ventilation simulator (from Wu et al. 1980)
Figure 3. Plan view of inner test room location and configureation (all dimensions are in meters)
 a |  b |
Figure 4. Sample images of room airflow from the PIV system (a) and air flow pattern (b) extracted from randomly located particle velocities
Figure 5. Interpolated flow pattern for Figure 4
Figure 6. Image of bubbles in the occupied zone of the test room with an air exchange rate fo 19.6 Air Changes per Hour (ACH)
Figure 7. Image of bubbles in the occupied zone in the test room with an air exchange rate of 15 ACH
Figure 8. Image of bubbles in the occupied zone in the test room with an air exchange rate of 9.6 ACH
Figure 9. The shifted (shift = 0.2 m/s) image of bubbles in the occupied zone of the test room with an air exchange rate of 15 ACH.
Figure 10. Flow velocity vector map directly extracted from Figure 9 (shift = 0.2 m/s)
Figure 11. Interpolated flow pattern for Figure 9