Determining 3D Flow Fields via Multi-camera Light Field Imaging

The overall aim of the following video is to present an overview of a three dimensional imaging technique that can yield a 3D velocity field. This is achieved by using calibrated cameras to gather the images required to sample the light field. As a second step, the light field is re parameterized, which produces a focal stack of images that form a 3D representation of the flow field.

Next, the focal stack is post-process using a cross correlation algorithm to obtain the 3D velocity field vectors. The results show a time resolved 3D flow field in the wake of a vibrating synthetic vocal fold model used as a test bed. Results are also shown for the technique applied to a bubble field.

Go ahead. The main advantage of this technique over existing methods is that we can measure in volumes that contain more particles, bubbles, or droplets. This method can provide insight into fluid flows and be extended to other applications, such as measurement of the shape of a flame and the role velocity plays in combustion and even to measurement of collective behavior of animal groups such as bird flocks.

Generally, individuals new to this technique will struggle because the amount of data can become overwhelming, but we think we've developed a cookbook for using this method.Okay. Visual demonstration of this method is critical because the camera and calibration setups are quite a bit different than when using a single camera approach. We'll be performing these experiments in Dr.Scott Thompson's BYU Biofluids lab with the help of his graduate student, Jesse Daley.

The first step is to determine the size of the measurement volume, as well as the temporal and spatial resolution required for investigating the fluid flow experiment. Here, the method will be used to perform 3D synthetic aperture particle image for loss symmetry on the airflow induced by a synthetic vocal fold. The measurement volume is 50 by 50 by 25 millimeter cubed, and the shortest timescales to be captured are 10 microseconds.

Next, estimate the optical density that will be present in the experiment in order to determine the number of cameras required to generate refocus images with good signal to noise ratio. Higher seeding densities require more cameras at this point in a particle image. For O symmetry experiments, the particles per pixel should also be calculated Mount cameras in an array configuration on a frame such that each camera can view the measurement volume from different viewpoints.

Next, set the spacing between the remaining cameras in the array. Spacing the cameras farther apart from each other improves the spatial resolution in the depth dimension at the cost of total resolvable depth. For data capture.

In viewing, attach the cameras to a central computer. Place a visual target such as a calibration grid in the center of the measurement volume. Use the image from the center camera of the array as a reference and move the entire array frame closer or farther from the measurement volume to achieve the desired magnification angle or cameras such that the visual target in the center of the measurement volume is approximately centered in each camera image.

With the apertures completely open on each camera lens, focus each camera on the visual target. Place a calibration target at the back of the measurement volume. Ensure that the target is in the view of each camera.

If it is not readjust the distance between the cameras and the measurement volume and or the camera spacing. Do the same with a calibration target at the front of the volume and iterate until the front and back are in view. In all cameras.

Close the aperture of each camera until the target is in focus. When located at any position within the measurement volume for each camera, additional lighting may be necessary with the aperture closed down. To begin, determine the appropriate method for illuminating the measurement volume based on the specific measurement method being applied to the flow field.

For this demonstration, a 1000 hertz double pulse laser is used. Use optical lenses to form the laser into a light volume that covers the measurement volume. Finally, when ready for data collection, be prepared to seed the volume with tracer particles suitable for particle image.Loc.

Symmetry measurements as described in the references. As a rule of thumb, an image density of 0.05 to 0.15 particles per pixel is appropriate for most experiments with eight or more cameras. For a fixed number of cameras, the particles per pixels decreases.

For larger volume depth dimensions. A critical step is calibration. This can be done with or without the tracer particles.

If using a multi-camera self calibration algorithm as in this demonstration, establish a reference coordinate system in the measurement volume. Here, the calibration grid is placed at the center of the vocal fold In a fixed orientation to the reference coordinate system, use an object with a known geometry as a calibration target. In this case, the calibration grid in the multi-camera self calibration algorithm or calibration target locations can be random except for the one precisely controlled.

That establishes the reference coordinate system In each camera, capture an image of the target in each location. Identify points on the target in each camera. For each image for self calibration, each identified point on the target must be located in the image made by each camera.

However, explicit location of the points in the reference coordinate system is only required for the points associated with the precisely located target. To acquire data for quantitative time resolved light field imaging, all cameras and illumination sources must be accurately synchronized. For this experiment, a programmed external pulse generator is used to trigger the camera exposures and illumination sequences.

Prepare for the collection of a large amount of data, including giving some thought to naming of data files. Begin experimental data capture by ensuring the tracer particles are flowing and initiating the camera capture and illumination sequence via the chosen triggering method. In order to produce a synthetically refocused volume for gathering data, generate a 3D focal stack.

To do this, define the spacing between focal planes and the overall refocusing depth in the refocused volume. As explained in the references, typically the focal plane is set to half the depth resolution and the total refocusing depth is governed by the region where all camera fields of view overlap. The focal planes will be perpendicular to the Z axis of the reference coordinate system.

Here we have a focal plane spacing of around 0.16 millimeters and a total refocus depth of 20 millimeters resulting in approximately 128 resolved seaplanes after processing, perform image pre-processing to improve background noise and accommodate differences in intensity between images. Establish transformations between each camera, image plane, and each synthetic focal plane. Re project images onto the synthetic focal planes.

Apply the scale and resample the images. This can be done within matlab. Given the plane to plane transformations apply either the additive or multiplicative synthetic aperture refocusing algorithm on each synthetic focal plane.

As a check, apply the refocusing to one plane of the calibration images to see if the reconstruction appears as expected. When the additive method is applied to one of the calibration planes at z equals 13.3 millimeters, the image comes into and outta focus as the focus stack is traversed from back to front. Finally, we demonstrate the focus at each calibration plane using the refocused images on the left and the image from the calibration grid from the central camera on the right.

After refocusing on all desired planes process, the images to remove noise caused by refocusing apply thresholding based on the intensity histograms of refocused images to retain in-focus particles. Next stack threshold images together to create a volume in a process called reconstruction. After reconstruction, quantitative data can be gathered from the volume.

An example of the high quality raw particle image for loss, symmetry images from a single camera is shown here. These images contain uniformly distributed particles appearing with a high contrast against the black background. Here is the result of an appropriately seeded and accurately calibrated experiment.

The synthetic aperture refocused image reveals in-focus particles on each depth plane from left to right are images at depths of minus seven millimeters, zero millimeters, and seven millimeters. To make use of the data requires a processing step known as reconstruction. In this case, intensity, thresholding is applied to retain in-focus particles on each depth plane.

The focal planes are then stacked to create a volume here. Images at the same depth are shown at two different times. The threshold volume can then be passed into interrogation volumes that contain an adequate number of particles for performing particle image velo symmetry.

This is an example of sample data gathered for the three dimensional vector field of the jet caused by synthetic vocal folds for several times steps. The left hand side shows an I asymmetric view of the entire 3D velocity field at each time. Step cuts of the XY plane at Z equal five millimeters is shown in the center cuts of the YZ plane.

At X equal 14 millimeters are shown at right at t equals zero milliseconds. The vocal fold is closed and very little velocity in the field is present. The largest speed in the jet at one millisecond moves in the positive wide direction and reduces in intensity from two to four milliseconds.

The fold closes at five milliseconds, reducing the jet speed and the cycle is repeated. This data represents the velocity field at a single snapshot in time as contrasted to the average that is usually presented. Another application of light field imaging is to bubbly flows.

Shown here is a bubbly field formed by the entrainment of air from a jet impinging on the surface of the water. Pausing the video at one time. Step allows refocusing through the image at different depth planes to see the bubbles come in and outta focus.

This still image shows from left to right, the raw image of a bubbly flow field from the camera array and refocused images at depths of minus 10 millimeters, zero millimeters, and 10 millimeters. The circle highlights a bubble that lies on the minus 10 millimeter depth plane and disappears from view on the other planes Once mastered, calibration and data capture can typically be performed in about four hours, and synthetic capture refocusing can be performed in about 12 hours while performing this procedure. It's important to be very organized as there are many steps in a lot of data that are collected.

Following this procedure, the rich data sets can be interrogated for physical insights into several questions such as what are the bubble size distributions in multi-phase flows? This technique will pave the way for researchers in fields such as physical biology, where they can study the fluid dynamics of butterfly flight or the three dimensional structure of bird flocking. After watching this video, you should have a pretty good understanding of how to set up cameras for light field imaging accurately calibrate them, perform synthetic aperture on the images in the software, and utilize the volumetric data for further processing.

For sample codes, data sets and tutorial information, please visit our website. Don't forget that working with Tad Truscott can be extremely hazardous and always take full precautions, such as wearing body armor when working in his lab.