From a DJI solar flight to a review-ready report
By 4Tech RealVision. Published . Updated .
What files do I need for a solar inspection?
Start with original radiometric thermal JPEGs and matching RGB photos from one solar flight. Keep the original metadata. Image overlap, capture conditions, and camera health affect suitability; model support alone does not establish compatibility. Tell us your drone model and site type. We confirm suitability and arrange beta access by email.
Captures processed end to end include DJI Matrice 4T, Mavic 3T (M3T / M3TA), and Mavic 2 Enterprise Advanced. The Franklin M2EA capture validates the pipeline only; a confirmed sensor fault makes its thermal data unsuitable for calibration validation.
How should I prepare my files before requesting beta access?
Keep one flight’s original thermal and RGB files together, with their metadata intact. The beta form asks about your team and dataset; it does not upload the imagery. Tell us the drone model, the kind of solar site, and whether both sets of originals are available. Keep a copy of the originals so you can return to the source evidence during review.
A colored thermal screenshot shows an appearance, but it is not a substitute for a radiometric JPEG containing temperature information. Likewise, a camera model on the processed-capture list does not establish that a particular flight has adequate overlap, usable metadata, or a healthy sensor. If you have only screenshots, missing RGB images, or files exported through another application, describe that limitation before arranging access. We can discuss suitability without presenting an incomplete dataset as ready to process.
What should I check in the maps before trusting a candidate location?
Check whether the map covers the area you intended to inspect and whether the candidate has enough location context to identify it. A completed map is not a promise that every panel was captured or assigned correctly. Compare the RGB and thermal views where available, then inspect the candidate’s source evidence. Keep uncertainty visible if the location is ambiguous or the candidate is unassigned.
For example, imagine a candidate near the edge of a mapped row. This is an illustrative review situation, not a result from a customer flight. A colored marker alone would not justify telling a technician which panel to replace. Review the available crop and source-image context, note the location uncertainty, and choose whether field verification or a not-assessable decision is appropriate. The useful output is an evidence-backed next step, not a more confident label.
How do I review candidate anomalies?
Review each candidate against the available maps and source evidence before deciding what it means. Heliovue builds georeferenced RGB and thermal maps and flags candidate hotspots and PID-like thermal patterns. Review Priority and Monitor queues help focus the review. Coverage and panel assignment depend on the capture; some locations remain ambiguous or unassigned.
For runs with complete candidate evidence, compare thermal and RGB map crops with source-image context. Record whether to retain or dismiss a candidate, request field verification, or mark it not assessable. Older captures may need reprocessing to enable evidence review.
What do the review decisions mean?
A review decision records what an inspector concludes from the available evidence; the automatic priority is only a starting point. For runs with complete evidence, the workflow lets you retain or dismiss a candidate, request field verification, or mark it not assessable, with a review note. A retained candidate still needs the level of verification appropriate to the job.
- Use the source context to explain why a candidate deserves attention.
- Record why a candidate was dismissed instead of treating an empty final list as proof of a healthy site.
- Identify what a field check needs to resolve when imagery is insufficient.
- Keep a not-assessable outcome visible when the evidence cannot support a conclusion.
What can I export after reviewing a flight?
You can generate an inspection PDF and export eligible candidate catalogs; available outputs depend on the capture. The IEC 62446-3-styled PDF brings together candidate findings, thermal imagery, review priorities, and available location information. Review the evidence and limitations before sharing it. Export eligible catalogs as CSV or GeoJSON, download available evidence archives, and create manual follow-up cases from eligible reviewed candidates.
These are automated screening results, not a diagnosis or a fault-free verdict. The report is not IEC certification. Field verification and a qualified inspector’s judgment remain part of the work.
What should I check before sharing the PDF?
Read the exported report against your reviewed evidence and intended inspection scope. Check that the locations and imagery are understandable, that uncertainties are explained, and that any requested follow-up is clear. Do not interpret the report’s format as a certification or the absence of candidates as proof that no fault exists.
You can inspect the three-page illustrative sample PDF before requesting beta access. It uses fictional data and schematic evidence to show a format; it is not a real inspection, a production export, or evidence of detection performance. Your own compatible dataset is needed to evaluate the actual review workflow and available outputs.
Where are images stored, and what if an older capture lacks evidence?
Product uploads and generated outputs are stored and processed using Google Cloud and Firebase, with organization access depending on member roles. Read the data-handling summary before sharing a dataset and confirm that you have authority to provide the files. You retain ownership of imagery and processing outputs; access, export, or deletion requests go through the published support contact.
Older captures may need reprocessing to enable evidence review. A missing archive or source crop is a limitation to investigate, not evidence that a candidate is false. Check what is available for that run before deciding whether the output is sufficient for your client’s job. Processing time and coverage depend on the dataset; the beta does not promise a turnaround or a fixed amount of time saved.
What happens during the first-flight beta?
We’ll help process one compatible dataset free and review the results with you. You keep the deliverables; we ask for one 20-minute feedback call.
Tell us about your first flight, read the beta questions and answers, or compare this workflow with reporting by hand.