Turning image quality into a decision, not a delay
At PLACE, we spend a lot of time solving problems that only show up once you’re actually in the field. One of those is image quality.
Drone missions can generate thousands of images in a single run. In BVLOS operations especially, UAVs often fly above low cloud layers, resulting in imagery that’s partially or fully compromised by haze, fog, or atmospheric conditions. Manually reviewing and sorting those images is slow, inconsistent, and can mean discovering too late that an entire flight needs to be re-flown.
So we built something better.
The Challenge
Assessing image quality at scale has traditionally been a manual, time-consuming task. Operators are left making subjective calls under time pressure, with real consequences for mission timelines, costs, and data quality.
Our Solution
We developed an Automated Drone Image Haze Analysis System that evaluates aerial imagery quality in minutes, not hours. Using computer vision, the system automatically analyzes atmospheric conditions across entire missions and delivers clear, objective reports to support fast re-flight decisions.
What it does
- Processes 15–50 images per second using CPU-based parallel processing
- Applies multi-factor haze scoring (contrast, brightness, saturation)
- Generates automated CSV reports per mission
- Supports batch processing across multiple flight folders
- Runs on standard field hardware — no GPU required
Why it matters
- 3–5× faster than manual review
- Fully automated, no subjective judgment
- Consistent, data-driven quality assessment
- Enables immediate decisions on whether and where to re-fly
Under the hood
Built in Python and OpenCV, the system uses parallel processing via ThreadPoolExecutor and quarter-resolution optimization for speed, while maintaining analytical accuracy through HSV color-space analysis.
Proven in the field
This tool proved its value during a recent mapping mission in the Kingdom of Tonga, where changing atmospheric conditions and BVLOS operations made real-time quality assessment critical. The system allowed our team to quickly identify compromised imagery, plan targeted re-flights, and significantly improve both mission efficiency and final data quality.
This is the kind of innovation we care about at PLACE: practical, field-tested, and focused on real problems.
Explore the code: GitHub Repository
We are incredibly proud of the team for building something that makes our work faster, smarter, and more reliable — and for proving that meaningful innovation often starts with the unglamorous problems no one else wants to tackle.
Already a member of the PLACE Community? Access our PLACE Data here.
Not yet a PLACE Member, but interested in using the data? Contact us or join today, and check out the available imagery via our STAC.