Sid Sharma.CV
Home / Projects / Software and modeling tools
A seeded random-dot background that the app generates. Its seed makes it reproducible, and it exports as PNG, vector PDF or HTML
Inside the app: the Learn tab, the physics lesson with typeset equations, and the displacement inspector, which shows the horizontal dot shift between two flow states

BOScope: Background-Oriented Schlieren on a Phone

Prototype software

An iPhone and Mac app that I built for fun. It turns a phone camera into a background-oriented schlieren (BOS) system. The app generates a random-dot background and captures a reference frame and a flow frame through it. It then cross-correlates the two frames to find the apparent dot displacement that density gradients cause. Examples are the plume above a candle or a heat gun.

Scope
A prototype with no validation yet. The synthetic-shift and physical accuracy checks are not complete, and the current estimates have detectable bias.
Methods & tools
Swift · SwiftUI · AVFoundation (locked focus, exposure and white balance, plus RAW companions) · Accelerate / Metal correlation · ARKit LiDAR · iOS and macOS targets
Links
Private · not released
Project breakdown

WhatSeeing density with a background

Schlieren imaging shows density gradients in a transparent flow. Background-oriented schlieren does this without mirrors or knife edges. A camera photographs a textured background through the flow, and light bent by the gradients shifts the apparent position of the texture. The measured shift gives the deflection. With the geometry, the deflection then gives the gradient.

\[n-1=K\rho,\qquad \varepsilon_y=\frac{1}{n_0}\int\frac{\partial n}{\partial y}\,dz,\qquad \Delta y\approx Z_D\,M\,\varepsilon_y\]

The Gladstone–Dale relation, the ray deflection through the flow, and the image displacement that it causes for a background distance \(Z_D\) and magnification \(M\).

HowFrom dots to a displacement field

  • Generates seeded, reproducible random-dot targets (fine, balanced or coarse) with corner codes for detection.
  • Captures a baseline frame and a flow-on frame with focus, exposure and white balance locked, or selects frames from video at up to 240 fps.
  • Estimates a two-component, sub-pixel displacement field by local normalized cross-correlation.
  • Shows signed Δx and Δy, magnitude and quality maps, vectors and a grayscale schlieren image.
  • Exports everything (images, CSV vectors, JSON metadata) as one archive.
  • Optionally uses LiDAR depth to check the distances that set the sensitivity.
The classical-schlieren view: filtered intensity, edges and the difference from the reference frame. You can export these with the raw fields.
Fig. 1 — The classical-schlieren view: filtered intensity, edges and the difference from the reference frame. You can export these with the raw fields.
\[R(\Delta\mathbf x)=\frac{\sum_{\mathbf x}\left[I_0(\mathbf x)-\bar I_0\right]\left[I_1(\mathbf x+\Delta\mathbf x)-\bar I_1\right]}{\sqrt{\sum\left[I_0-\bar I_0\right]^2\sum\left[I_1-\bar I_1\right]^2}}\]

Normalized cross-correlation between interrogation windows of the reference and flow images. The sub-pixel peak gives the local displacement.

ResultsScope of the prototype

The full pipeline runs on the phone: target, capture, correlation, inspection and export. But it does not guarantee accuracy. The synthetic-shift, physical target-motion and ray-angle checks are not complete, so the app labels its results as prototype processing. Density and temperature reconstruction, 3-D reconstruction and velocimetry are not implemented.

The Learn tab is also a short course on schlieren physics. It includes a gallery from the MECH 489 Z-type schlieren lab that started the project.