I Couldn't See Stars from My Balcony. So I Built a Patent-Pending System to Measure Light Pollution.

The story of how a Class 11 student in Gurgaon turned a simple question into SkyQI -- a free platform that turns any smartphone into a sky quality meter

I couldn't see stars from my balcony in Gurgaon. That question -- why can't I see stars? -- turned into months of research, thousands of lines of code, and eventually, a patent application filed with the Indian Patent Office.

My name is Suhani Gupta. I'm a Class 11 student at Amity International School, Sector 46, Gurgaon, Haryana. And this is the story of how I built SkyQI.

The Question That Started Everything

I grew up in one of India's fastest-growing cities. Gurgaon has gone from farmland to a forest of glass towers in barely two decades. The malls, the offices, the highways -- they all came with light. So much light that the night sky turned from black to a permanent grey-orange haze.

I didn't really notice it until I visited a relative's home in a rural part of Rajasthan a few years ago. That night, I walked outside and looked up. I could see thousands of stars. The Milky Way was right there, a bright band stretching overhead. I had only ever seen that in photographs.

Back in Gurgaon, I looked up again. Maybe a dozen stars, if I was lucky.

That contrast stuck with me. I started reading about light pollution and learned a statistic that I still find hard to believe: roughly 80% of Indians cannot see the Milky Way from where they live. It's not that the stars have gone anywhere. It's that we've drowned them out with our own light.

I wanted to measure it. How bad is the light pollution above my house, exactly? And how does it compare to other places?

Why Existing Solutions Failed Me

The standard tool for measuring sky brightness is a Sky Quality Meter, or SQM -- a handheld device made by a Canadian company called Unihedron. It gives you a precise reading in magnitudes per square arcsecond. The problem: these devices cost between $150 and $400. For a student, that was out of the question.

I looked at satellite data next. NASA's VIIRS satellite measures light radiance from orbit and the data is freely available. But there's a fundamental problem: satellites measure light going up from the ground. What I wanted to know was the sky quality looking down from above -- or rather, looking up from where I stand. These are related but not the same thing. A city with good lighting design can emit less light upward than a poorly-lit smaller town.

Then there's visual estimation using the Bortle scale, where you look at the sky and compare what you see to a set of descriptions (Can you see the zodiacal light? Can you resolve the Milky Way into individual stars?). This requires astronomical experience I simply didn't have.

Every existing method was either too expensive, too indirect, or too dependent on expertise. I kept thinking: everyone carries a camera in their pocket. Why can't we use phone photos to measure sky quality?

Building the Algorithm

In May 2025, I started teaching myself image processing. I had some programming background, but working with pixel-level image data was new territory. The early attempts were rough. My first version just measured average brightness of a photo and mapped it to a Bortle class. It was wildly inaccurate -- a photo of a cloudy sky and a photo of a light-polluted sky looked the same to it.

The key insight came after weeks of failed approaches: no single measurement from an uncalibrated phone camera is reliable enough on its own. Phone cameras vary enormously -- different sensors, different processing pipelines, different exposure settings. Any one indicator could be thrown off.

So instead of relying on one measurement, I designed a system that uses three independent indicators and requires them to agree:

Star density ratio -- how many point-like bright objects appear relative to the image size. Dark skies have more visible stars.

Brightness gradient ratio -- how the brightness changes from the center of the sky down to the horizon. In light-polluted areas, the horizon glows much brighter than the zenith.

Calibrated SQM -- an estimated sky brightness in magnitudes per square arcsecond, calculated from pixel luminance values using the ITU-R BT.709 standard.

The cross-validation gate is what makes it work. If the three indicators disagree significantly, the system flags the measurement as unreliable rather than guessing. This catches clouds, sensor noise, unusual atmospheric conditions, and images that aren't actually night sky photos. It's also what allows the system to work across wildly different phone cameras without requiring device-specific calibration.

I spent the summer of 2025 building, testing, breaking, and rebuilding. The algorithm went through more iterations than I can count. There were weeks where I thought the whole approach was fundamentally flawed. The star detection would pick up hot pixels as stars. The gradient analysis would get confused by the Moon. The SQM calculation would give absurd values for overexposed photos. Each problem had to be solved, and solving one often broke something else.

By September 2025, I had a working platform.

The Validation

An algorithm is only as good as its validation. I tested SkyQI against 1,074 ground truth measurements sourced from professional SQM devices, Globe at Night visual observations, and VIIRS satellite data covering 60,681 grid points across India.

The result: 65.4% accuracy within one Bortle class.

That number deserves honest context. It's not perfect. Professional SQM devices are far more precise. But SkyQI isn't trying to replace SQM devices -- it's trying to do something SQM devices can't: scale to millions of people who will never buy a $200 instrument but who do carry smartphones.

For citizen science at scale, being approximately right across thousands of measurements is more valuable than being precisely right in a handful of locations.

Filing the Patent

In February 2026, I filed Indian Patent Application No. 202611013909, titled "Self-Calibrating Multi-Parameter Sky Quality Measurement System with Cross-Validation for Uncalibrated Consumer Imaging Devices."

I won't pretend the filing process was simple. Patent language is its own world -- every word carries legal weight, and the difference between a strong claim and a weak one can be a single phrase. But the core idea was clear: a system that uses multiple independent sky quality indicators with cross-validation to produce reliable measurements from uncalibrated consumer cameras. That specific combination -- the multi-parameter approach with the cross-validation gate -- is what I believe is genuinely novel.

Seeing my work formalized as intellectual property was surreal. Five months earlier, I was writing my first star detection function. Now I was filing claims with the Indian Patent Office.

What SkyQI Does Today

SkyQI is a free platform. You take a photo of the night sky with your phone, upload it, and get instant analysis: your SQM value, Bortle class, star count, confidence level, and a breakdown of what the algorithm detected.

There's a global map where you can see measurements from other users, overlaid with VIIRS satellite data so you can compare ground-level and satellite perspectives. The platform is built with React and Node.js on the frontend and backend, TypeScript throughout, Flutter for the mobile app, and Sharp for image processing. Everything runs on a single server.

People in more than seven countries have used it. I've seen measurements come in from places I've never heard of -- small towns, national parks, university campuses. Each data point fills in a piece of the global light pollution picture.

What I've Learned

The most important thing I've learned is that the best solutions come from problems you personally experience. I didn't set out to build a platform or file a patent. I just wanted to know how dark my sky was. Every decision I made -- to use phone cameras, to make it free, to build a web platform instead of requiring a dedicated app -- came from my own constraints as a student without funding or institutional support.

I've also learned that you don't need expensive equipment to do real science. The phone in your pocket is a sensor array -- camera, GPS, accelerometer, magnetometer. The question is whether you can write software smart enough to extract useful data from imperfect hardware. That's what SkyQI attempts.

And perhaps most importantly: technology should make science accessible, not gatekeep it. If measuring light pollution requires a $400 device and a PhD in astronomy, then only a handful of people will ever do it. If it requires a free website and a phone camera, then anyone can.

What's Next

I want to make light pollution visible to everyone. Not just as a concept -- as a number, tied to their specific location, that they can track over time.

I'm planning workshops in schools to get students measuring their own sky quality. I want to build partnerships with researchers who study light pollution and need ground-level data at a scale that traditional instruments can't provide. And I want to keep improving the algorithm -- every new measurement is a data point I can learn from.

The long-term vision is simple: turn every smartphone into a scientific instrument. Not a toy that approximates science, but a real tool that produces data good enough for research.

Try It

If you've ever wondered how dark your night sky really is -- now you can find out.

Visit skyqi.in, point your phone at the sky, take a photo, and upload it. In a few seconds, you'll know exactly where your sky stands on the Bortle scale.

And if you do try it, you'll be contributing to a growing dataset that maps light pollution from the ground up -- one photo at a time.