At PV Magic, the accuracy of our solar panel analytics begins with one critical step: high-quality data acquisition. To ensure precise defect detection and analysis, we’ve outlined comprehensive guidelines for capturing, organizing, and submitting drone imagery of photovoltaic (PV) systems.
In this post, we’re summarizing the essential rules that help our system deliver the best results — whether you’re a drone operator, data provider, or solar farm manager.
1. Flight and Image Acquisition Best Practices
– Use path planning, not manual control — consistent flight paths improve overlap and coverage.
– Maintain a perpendicular camera angle relative to the PV panels for maximum resolution.
– Altitude and angle stability: Avoid altering drone height or gimbal angle mid-flight.
– Overlap matters: Ensure at least 60% image overlap between shots.
– Use RTK GPS for drone position accuracy within 10 cm.
– Fly when the weather helps you, not hinders you:
– Wind speed should stay below 28 km/h.
– Cloud cover should not exceed 2 okta (light cumulus clouds).
2. Quality Assurance Requirements
– Minimum 5×5 pixels per cell resolution on all modules.
– Camera angle should be as close to 90° to the PV module surface as possible.
– Images that are blurry, overexposed, or poorly framed are not usable.
3. Proper File Naming and Structure
– Naming format: <OPTIONAL_PREFIX>_<DATETIME>_<IMG_NUM>_<TAG>.JPG
Example: DJI_20240730123933_0001_T.JPG
– TAG = “T” for thermal, “V”/”Z”/”W” for RGB
– Folder structure: Group each sector in its own folder with RGB/thermal image pairs.
– Optional Checkpoints: Use <CHECKPOINT> to mark row starts/ends for easier orientation.
4. Include Metadata
We rely on embedded metadata for proper panel alignment and georegistration. This includes:
– GPS coordinates (latitude, longitude, altitude)
– Orientation and camera gimbal angles
– Thermal data (DJI format supported)
5. What Makes an Image Unusable?
Some examples of data we cannot process:
– Images taken from too high up → insufficient resolution
– Drone/camera not aligned with panel rows
– Farms without defined rows
– Overlapping modules or unclear regions of interest
6. Optional but Recommended: Overview Images
Capturing a full-sector high-altitude overview image helps with orientation and reporting. While not required, it’s a valuable addition to any dataset.
Final Note
While these guidelines maximize the chances of successful analysis, not every dataset that meets them will be processable — and occasionally, data that doesn’t fully comply may still work. Each submission is reviewed case by case by our processing team.
By following these recommendations, you’re helping us — and your team — get the most accurate and actionable insights from your solar inspections.
For questions or to upload your dataset, visit www.pvmagic.com or contact us at magic@pvmagic.com.