Statistical investigation of errors in particle image velocimetry

The accuracy of the particle image velocimetry technique was investigated using synthetic images having known characteristics. Algorithms were developed to extract two-dimensional velocity information by tracking particles between successive frames of a movie automatically without operator assistance. This allowed to parametrically investigate the influence of the various parameters (image contrast, image noise, particle density, distribution of sizes of particles and particle displacement between frames) on the accuracy of the technique. It was found that as long as the images have a good contrast, particle locations can be determined with sub-pixel accuracy and particle velocities can be determined within a few percent.     

Digital particle image velocimetry

Digital particle image velocimetry (DPIV) is the digital counterpart of conventional laser speckle velocitmetry (LSV) and particle image velocimetry (PIV) techniques. In this novel, two-dimensional technique, digitally recorded video images are analyzed computationally, removing both the photographic and opto-mechanical processing steps inherent to PIV and LSV. The directional ambiguity generally associated with PIV and LSV is resolved by implementing local spatial cross-correlations between two sequential single-exposed particle images. The images are recorded at video rate (30 Hz or slower) which currently limits the application of the technique to low speed flows until digital, high resolution video systems with higher framing rates become more economically feasible. Sequential imaging makes it possible to study unsteady phenomena like the temporal evolution of a vortex ring described in this paper. The spatial velocity measurements are compared with data obtained by direct measurement of the separation of individual particle pairs. Recovered velocity data are used to compute the spatial and temporal vorticity distribution and the circulation of the vortex ring.     

A generalized processing technique in digital particle image velocimetry with direct estimation of velocity gradients

A technique is proposed for the processing of digital particle image velocimetry (PIV) images, in one single step providing direct estimates of fluid velocity, out-of-plane vorticity and in-plane shear rate tensor. The method is based on a generalization of the standard PIV cross-correlation technique and substitutes the usual discrete cross-correlation of image pairs with a correlation of interpolated two-dimensional image intensity functions, being subject to affine transformations. The correlation is implemented by using collocation points, on which image intensity values are interpolated. The resulting six-dimensional correlation function is maximized using a general purpose optimization algorithm. The use of the method is demonstrated by application to different types of synthetically generated image pairs constructed with known particle displacement functions. The resulting errors are assessed and compared with those of a representative standard PIV method as well as with those of the present technique using no differential quantities in the search of the peak location. The examples demonstrate that significant improvements in accuracy can be obtained for flow fields with regions containing strong velocity gradients.     

A hybrid digital particle tracking velocimetry technique

A novel approach to digital particle tracking velocimetry (DPTV) based on cross-correlation digital particle image velocimetry (DPIV) is presented that eliminates the need to interpolate the randomly located velocity vectors (typical of tracking techniques) and results in significantly improved resolution and accuracy. In particular, this approach allows for the direct measurement of mean squared fluctuating gradients, and thus several important components of the turbulent dissipation. The effect of various parameters (seeding density, particle diameter, dynamic range, out-of-plane motion, and gradient strength) on accuracy for both DPTV and DPIV are investigated using a Monte Carlo simulation and optimal values are reported. Validation results are presented from the comparison of measurements by the DPTV technique in a turbulent flat plate boundary layer to laser Doppler anemometer (LDA) measurements in the same flow as well as direct numerical simulation (DNS) data. The DPIV analysis of the images used for the DPTV validation is included for comparison. Received: 29 August 1994/Accepted: 31 May 1996     

3D particle streak velocimetry by defocused imaging

Particle streak velocimetry (PSV) has become one of the important branches of flow filed measurements. It extracts velocity information from particle trajectories captured by single frame long exposure images. Since the defocus of moving particle is inevitable during a long exposure time and under a large magnification, a novel three-dimensional (3D) velocity measurement method named defocusing particle streak velocimetry (DPSV) is proposed in this paper. On the one hand, an extension from two-dimensional (2D) to 3D velocity measurement with a monocular system is carried out. The depth information of the particle, which reflects the position in the third dimension, is indicated by the defocusing degree (characteristic parameter σ) of the particle images. The variation of σ along the trajectory is recognized by surface fitting of the gray value distribution of particle images, assuming that σ varies linearly along the trajectory. On the other hand, based on the linear fitting for the straight trajectory, an arc fitting model is developed for curved trajectories which are commonly captured in turbulent flow. The relationship between σ and the particle depth position z is experimentally calibrated using a LED light and a diaphragm. Finally, the DPSV method is verified in a submerged jet flow field as well as in a microchannel flow field to obtain the 3D velocity field with single monocular system.     

New insights into particle image velocimetry data using fuzzy-logic-based correlation/particle tracking processing

Digital particle image velocimetry (DPIV) data processing has been developed to the point where DPIV image data are processed via auto- or cross-correlation techniques in near real time and the results are displayed on screen as they are processed. Correlation techniques are highly desirable, since they provide velocity measurements on a regular grid, which are readily comparable to CFD predictions of the flow field. In high-speed flows, particle lag effects are always of concern; however, the correlation operation does not provide any means for minimization or elimination of systematic errors in the recorded particle image data. In this paper, we present a combined correlation processing/particle tracking technique providing “super-resolution” velocity measurements. Fuzzy-logic principles are employed to maximize the information recovery in the correlation operation and to determine the correct particle pairings in the tracking operation. The combined correlation/particle tracking technique is applied to DPIV data obtained in the diffuser region of a high-speed centrifugal compressor producing velocity vector maps with an average density of 6 vectors/mm². Inspection of the particle tracking results revealed large particles that were not following the flow. Using preknowledge of the flow field, the biased velocity estimates arising from large particles in the flow were removed, thereby improving the accuracy of the measurements. Received: 21 October 1999/Accepted: 19 August 2000     

3D scanning particle tracking velocimetry

In this article, we present an experimental setup and data processing schemes for 3D scanning particle tracking velocimetry (SPTV), which expands on the classical 3D particle tracking velocimetry (PTV) through changes in the illumination, image acquisition and analysis. 3D PTV is a flexible flow measurement technique based on the processing of stereoscopic images of flow tracer particles. The technique allows obtaining Lagrangian flow information directly from measured 3D trajectories of individual particles. While for a classical PTV the entire region of interest is simultaneously illuminated and recorded, in SPTV the flow field is recorded by sequential tomographic high-speed imaging of the region of interest. The advantage of the presented method is a considerable increase in maximum feasible seeding density. Results are shown for an experiment in homogenous turbulence and compared with PTV. SPTV yielded an average 3,500 tracked particles per time step, which implies a significant enhancement of the spatial resolution for Lagrangian flow measurements.     

Effect of resolution on the speed and accuracy of particle image velocimetry interrogation

Particle image velocimetry incorporates a process by which an image of a flow field, bearing double images of seeding particles, is analyzed in small regions called “interrogation spots.” Each spot is imaged onto a photodetector array whose digitized output is evaluated computationally using the auto-correlation technique. This paper examines the effects of resolving the spot using arrays of various resolutions, motivated primarily by a gain in speed. For this purpose, two specially created test photographs representing (i) uniform flow and (ii) solid body rotation, were interrogated using array sizes ranging from 32 × 32 to 256 × 256. Each reduction in resolution by a factor of two gains a factor of four in interrogation speed, but this benefit is counteracted by a loss in accuracy. The particle image diameter strongly influences accuracy through two distinct error mechanisms. When the particle image is small compared to the pixel size, mean bias error becomes significant due to finite numerical resolution of the correlation function. Conversely, when the particle image is large, random error due to irregularities in the electronic images predominates. The optimum image size, therefore, lies not at either extreme but at an intermediate value such that the particle image is small in an absolute sense, and yet large relative to the pixel size. A version of this paper was presented at the 12th Symposium on Turbulence, University of Missouri-Rolla, 24–26 September 1990     

High-precision sub-pixel interpolation in particle image velocimetry image processing

The task of image interpolation and re-sampling for particle image velocimetry (PIV) is investigated, which is used for window shifting with sub-pixel accuracy and image or window deformation. A new interpolation scheme based on a Gaussian filter is introduced and compared with commonly used and widely accepted interpolation techniques in terms of the achievable root mean square deviation of the displacement estimates.     

A method for automatic estimation of instantaneous local uncertainty in particle image velocimetry measurements

The uncertainty of any measurement is the interval in which one believes the actual error lies. Particle image velocimetry (PIV) measurement error depends on the PIV algorithm used, a wide range of user inputs, flow characteristics, and the experimental setup. Since these factors vary in time and space, they lead to nonuniform error throughout the flow field. As such, a universal PIV uncertainty estimate is not adequate and can be misleading. This is of particular interest when PIV data are used for comparison with computational or experimental data. A method to estimate the uncertainty from sources detectable in the raw images and due to the PIV calculation of each individual velocity measurement is presented. The relationship between four error sources and their contribution to PIV error is first determined. The sources, or parameters, considered are particle image diameter, particle density, particle displacement, and velocity gradient, although this choice in parameters is arbitrary and may not be complete. This information provides a four-dimensional “uncertainty surface” specific to the PIV algorithm used. After PIV processing, our code “measures" the value of each of these parameters and estimates the velocity uncertainty due to the PIV algorithm for each vector in the flow field. The reliability of our methodology is validated using known flow fields so the actual error can be determined. Our analysis shows that, for most flows, the uncertainty distribution obtained using this method fits the confidence interval. An experiment is used to show that systematic uncertainties are accurately computed for a jet flow. The method is general and can be adapted to any PIV analysis, provided that the relevant error sources can be identified for a given experiment and the appropriate parameters can be quantified from the images obtained.     

Template matching for improved accuracy in molecular tagging velocimetry

In 2D molecular tagging velocimetry (MTV), tags are written into a fluid flow with a laser grid and imaged at discrete times. These images are analyzed to calculate Lagrangian displacement vectors, often by direct cross correlation. The cross correlation method is inherited from particle imaging velocimetry, where the correlated images contain a random pattern of particles. A template matching method is presented here which takes advantage of the known geometry of laser written tag grids in MTV to achieve better accuracy. Grid intersections are explicitly located in each image by correlation with a template with several linear and rotational degrees of freedom. The template is a continuous mathematical function, so the correlation may be optimized at arbitrary sub-pixel resolution. The template is smooth at the spatial scale of the image noise, so random error is substantially suppressed. Under typical experimental conditions at low imaging resolution, displacement uncertainty is reduced by a factor of 5 compared to the direct cross correlation method. Due to the rotational degrees of freedom, displacement uncertainty is insensitive to highly deformed grids, thus permitting longer delay times and increasing the relative accuracy and dynamic range of the measurement. In addition, measured rotational displacements yield velocity gradients which improve the fidelity of interpolated velocity maps.     

Analysis and treatment of errors due to high velocity gradients in particle image velocimetry

This paper deals with errors occurring in two-dimensional cross-correlation particle image velocimetry (PIV) algorithms (with window shifting), when high velocity gradients are present. A first bias error is due to the difference between the Lagrangian displacement of a particle and the real velocity. This error is calculated theoretically as a function of the velocity gradients, and is shown to reach values up to 1 pixel if only one window is translated. However, it becomes negligible when both windows are shifted in a symmetric way. A second error source is linked to the image pattern deformation, which decreases the height of the correlation peaks. In order to reduce this effect, the windows are deformed according to the velocity gradients in an iterative process. The problem of finding a sufficiently reliable starting point for the iteration is solved by applying a Gaussian filter to the images for the first correlation. Tests of a PIV algorithm based on these techniques are performed, showing their efficiency, and allowing the determination of an optimum time separation between images for a given velocity field. An application of the new algorithm to experimental particle images containing concentrated vortices is shown.     

Measurement of fully-developed turbulent pipe flow with digital particle image velocimetry

A new and unique high-resolution image acquisition system for digital particle image velocimetry (DPIV) in turbulent flows is used for the measurement of fully-developed turbulent pipe flow at a Reynolds number of 5300. The flow conditions of the pipe flow match those of a direct numerical simulation (DNS) and of measurements with conventional (viz., photographic) PIV and with laser-Doppler velocimetry (LDV). This experiment allows a direct and detailed comparison of the conventional and digital implementations of the PIV method for a non-trivial unsteady flow. The results for the turbulence statistics and power spectra show that the level of accuracy for DPIV is comparable to that of conventional PIV, despite a considerable difference in the interrogation pixel resolution, i.e. 32 × 32 (DPIV) versus 256 × 256 (PIV). This result is in agreement with an earlier analytical prediction for the measurement accuracy. One of the advantages of DPIV over conventional PIV is that the interrogation of the DPIV images takes only a fraction of the time needed for the interrogation of the PIV photographs.     

Application of numerical and optical evaluation schemes for particle image velocimetry

Application of particle image velocimetry (PIV) techniques for measurement of fluid velocities typically requires two steps. The first of these is the photography step in which one or more exposures of a particle field are taken. The second step is the evaluation of the particle pattern and production of appropriate velocities. Each of these steps involves optimization which is usually specific to the experiment being conducted and there is significant interaction between photographic parameters and evaluation characteristics.Among the various evaluation techniques suggested for analysis of PIV images is the evaluation of the scattered interference pattern (Young's fringes) by numerical Fourier transform. An alternative to the numerical calculation of the Fourier transform of the Young's fringes has been suggested, using a modified liquid crystal television as an optical correlator to allow the transform to be performed optically. Both transform techniques are affected by the quality of the input function, specifically the Young's fringes.This paper will compare the performance of optical and numerical Fourier transform analysis of Young's fringes using speckle images. The repeatability and an estimate of the accuracy of the particle displacement will be shown for each method. A brief examination of the effects of small particle number density of PIV evaluation will also be presented. Finally, for a small part of an actual unsteady flow, the optical and numerical Fourier transform analysis methods will be compared.     

Optimal subpixel interpolation in particle image velocimetry

It is shown that among the different techniques for particle image velocimetry subpixel interpolation, only the "sinc"-kernel creates an optimal result in that it completely suppresses spurious spectral sidelobes. An efficient method is introduced for the computation of the subpixel-accurate correlation peak position without any systematic errors. A connection is made with the kernel-dependent observation of the peak-locking phenomenon.     

Estimation of complex air–water interfaces from particle image velocimetry images

This paper describes a method for the estimation of the instantaneous air–water interface directly from particle image velocimetry (PIV) images of a laboratory generated air entraining turbulent hydraulic jump. Image processing methods such as texture segmentation based on gray level co-occurrence matrices are used to obtain a first approximation for the discrete location of the free surface. Active contours based on energy minimization principles are then implemented to get a more accurate estimate of the calculated interface and draw it closer to the real surface. Results are presented for two sets of images with varying degrees of image information and surface deformation. Comparisons with visually-interpreted surfaces show good agreement. In the absence of in-situ measurements, several verification tests based on physical reasoning show that the free surface is calculated to acceptable levels of accuracy. Aside from a single image used to tune the set of parameters, the algorithm is completely automated to process an ensemble of images representative of typical PIV applications. The method is computationally efficient and can be used to track fluid-interfaces undergoing non-rigid deformations.     

Twenty years of particle image velocimetry

The development of the method of particle image velocimetry (PIV) is traced by describing some of the milestones that have enabled new and/or better measurements to be made. The current status of PIV is summarized, and some goals for future advances are addressed.     

Integrating cross-correlation and relaxation algorithms for particle tracking velocimetry

An integrated cross-correlation/relaxation algorithm for particle tracking velocimetry is presented. The aim of this integration is to provide a flexible methodology able to analyze images with different seeding and flow conditions. The method is based on the improvement of the individual performance of both matching methods by combining their characteristics in a two-stage process. Analogous to the hybrid particle image velocimetry method, the combined algorithm starts with a solution obtained by the cross-correlation algorithm, which is further refined by the application of the relaxation algorithm in the zones where the cross-correlation method shows low reliability. The performance of the three algorithms, cross-correlation, relaxation method and the integrated cross-correlation/relaxation algorithm, is compared and analyzed using synthetic and large-scale experimental images. The results show that in case of high velocity gradients and heterogeneous seeding, the integrated algorithm improves the overall performance of the individual algorithms on which it is based, in terms of number of valid recovered vectors, with a lower sensitivity to the individual control parameters.     

Improving the dynamic range of particle tracking velocimetry systems

In applying a video-based particle image velocimetry (PTV) system in a complex fluid flow, it is common to find both regions of fast and slow moving flow intermixing-particularly in highly turbulent or reversing flows. When one attempts to track the movement of particles in such a flow with a wide velocity range (and hence, separation distance between particle images), resolution problems are encountered. Inability to cover a wide range of velocities is actually a limitation of PTV. A method is introduced here that extends the dynamic range of PTV when implemented on a video-based system. It combines the use of multiple frames and multiple exposures on a single frame. The method is subsequently verified by tracking dots painted on a spinning flat disc.     

Volume self-calibration for 3D particle image velocimetry

Planar self-calibration methods have become standard for stereo PIV to correct misalignments between laser light sheet and calibration plane. Computing cross-correlation between images from camera 1 and 2 taken at the same time, non-zero disparity vectors indicate rotational and translational misalignments relative to the coordinate system defined by a calibration plate. This approach works well for thin light sheets but fails for extended volumes recorded in 3D-PTV or tomographic PIV experiments. Here it is primarily necessary to correct calibration errors leading to triangulation errors in 3D-PTV or in degraded tomographic volume reconstruction. Tomographic PIV requires calibration accuracies of a fraction of a pixel throughout the complete volume, which is difficult to achieve experimentally. A new volumetric self-calibration technique has been developed based on the computation of the 3D position of matching particles by triangulation as in 3D-PTV. The residual triangulation error (‘disparity’) is then used to correct the mapping functions for all cameras. A statistical clustering method suitable for dense particle images has been implemented to find correct disparity map peaks from true particle matches. Disparity maps from multiple recordings are summed for better statistics. This self-calibration scheme has been validated using several tomographic PIV experiments improving the vector quality significantly. The relevance for other 3D velocimetry methods is discussed.