← All work

Remote Lab Vision

Undergraduate research · Oregon State University

Remote Lab Vision

2025 – 2026

Abstract

Two lab rigs let students run chemistry and chemical-engineering lab work through a computer, observing it through a camera. I benchmarked four cameras in LabVIEW on the resolution and frame rate each actually delivered and on their RGB intensity histograms, picked the OBSBOT Meet 2 4K, mounted it on aluminium extrusion and presented the work as a research poster.

Keywords remote labs · machine vision · LabVIEW · NI-IMAQdx · camera benchmark

Role
Undergraduate researcher
Programme
URSA Engage
With
Prof. Lucas Ellis, Stephen Stockton
Tools
LabVIEW, NI-IMAQdx

Question

Two lab rigs let students run chemistry and chemical-engineering lab work remotely, through a computer. They observe, measure and get results without being in the lab, which helps the students who cannot be there in person.

Observing from somewhere else means observing through a camera, and its picture has to arrive inside LabVIEW, beside the rig's controls control panel. The poster puts the objective in one line: “to establish a stable, high-resolution video feed directly into LabVIEW.”

A laptop showing the LabVIEW control panel: valve and column controls on the left, the camera's live view of the rig on the right
The LabVIEW control panel: the rig's controls on the left, the camera's live view on the right.

Which camera can hold that feed, and how would you know?

Rigs

Rig one is built on a steel panel: nitrogen and air valves, two mass-flow controllers, a control box, and two clear columns with probes in them rig one.

Rig one on a steel panel: nitrogen and air valves, two mass-flow controllers, a control box with a red switch, and two clear columns with probes
Rig one: N₂ and air valves, two mass-flow controllers, a control box, two clear columns with probes.

Rig two is a wall panel of sensor fittings, with its own control box and emergency stop, and a laptop on the bench below it rig two.

Rig two: a wall panel of blue sensor fittings above a control box with an emergency stop, and a laptop on the bench below
Rig two: a wall panel of sensor fittings, a control box with an e-stop, and a laptop.

The clip walks through both, as they stand in the lab walk-through.

Both rigs, walked through. No sound. 0:31

Method

I benchmarked four cameras: the OBSBOT Meet 2 4K, the Logitech C922, the GoPro Hero 8 and the Yi. For a test, the candidate camera was clamped to a stand and aimed at the rig test setup.

A candidate camera held by a clamp on a lab stand, aimed at the rig's two columns
The test setup: a candidate camera, clamped and aimed at the rig.

Each one fed LabVIEW through NI-IMAQdx, the module the poster names to “acquire and process real-time video streams” block diagram.

The LabVIEW block diagram on a laptop screen, including the NI-IMAQdx camera image capture loop
The block diagram, with the NI-IMAQdx camera capture loop.

Then the comparison: not the resolution and frame rate on each camera's box, but what reached LabVIEW, the “‘true’ resolution and frame rate (FPS) actually received”, in the poster's words objectives. The drivers were tested too, standard UVC plug-and-play against proprietary video streams.

The poster's Objectives panel: software integration, driver compatibility testing, performance analysis and hardware assembly
The method, as the poster's Objectives panel lists it.

Last, the picture itself. The poster sets the four cameras' views of the rig's columns side by side camera views, and compares them by RGB intensity histogram.

Four views of the rig's columns from the poster, labelled Yi Action Camera, OBSBOT Meet 2-4K, Logitech C922 Pro Stream and GoPro Hero 8 Black
The rig's columns through each camera, as labelled on the poster: Yi, OBSBOT, Logitech, GoPro.

Results

The histograms put pixel intensity, 0 to 255, against a logarithmic pixel count histograms. The poster's reading: “a broader and smoother distribution curve indicates a superior dynamic range.”

RGB pixel intensity histograms from the poster for the OBSBOT Meet 2-4K, Logitech C922 Pro Stream, GoPro Hero 8 Black and Yi Action Camera
RGB pixel intensity histograms for the four cameras, from the poster.

I picked the OBSBOT Meet 2 4K. From the poster: it “emerged as the most viable current solution”, with “the best balance of high-resolution imaging, system compatibility, and cost-efficiency” conclusion.

The poster's Conclusion and Future Efforts panel
Conclusion and Future Efforts, from the poster.

The other three fell short in two ways. The action cameras, GoPro and Yi, “failed entirely due to closed software ecosystems”; the Logitech webcam “lacked the necessary visual detail”.

I mounted the OBSBOT on aluminium extrusion, and the poster's method has the focal length calibrated “to the target object” at the rig.

Jakhangir standing at the rig, the camera feed of the rig live in LabVIEW on the monitor beside him
At the rig, with the camera feed live in LabVIEW.

Limits

The OBSBOT is a baseline. Its drawback, from the poster, is “the fixed lens architecture, which forces us to make manual physical adjustments” conclusion.

The next step the poster names is DSLR systems, “specifically Canon models (e.g., Rebel T5i or T8i)”, for interchangeable lenses and true optical zoom.

And the comparison covers four cameras. It chose among them; it does not rank every camera a remote lab could use.

Poster

I presented the work as a research poster, “Evaluation and Integration of a Vision System for Remote Learning Labs”, with Stephen Stockton and Lucas Ellis poster.

Jakhangir standing beside his research poster in a hall of posters, a man on its other side
With the poster.

Acknowledgements

URSA Engage undergraduate research at Oregon State University, 2025 – 2026, with Prof. Lucas Ellis and Stephen Stockton.