Introduction
Strayos’ Fragmentation AI is a powerful tool that leverages deep learning to
analyze muckpile images, providing accurate rock size distributions and blast
performance insights for mining operations. This Quality Assurance/Quality
Control (QA/QC) workflow guide ensures that customers collect high-quality
data, process it effectively, and interpret results reliably.
Pre‑Flight Best Practices
|
Task
|
Recommendation
|
Why
|
|
Timing
|
Capture 10–15 min after blast when dust settles.
Avoid long shadows
|
Clear edges, fewer
artifacts.
|
|
Altitude
|
20–40 m AGL (GSD ≤ 2 cm).
|
Detects ≥ 10 mm fragments.
|
|
Overlap
|
80 % front / 70 % side.
|
Fewer voids in point cloud.
|
Environmental Considerations
- Lighting Fly during optimal lighting conditions (e.g.,
mid-morning or early afternoon) to minimize shadows that can obscure rock
edges.
- Weather Avoid flying in rain, high winds, or fog, which can
affect image quality or drone stability.
- Calibration : Verify drone sensors (e.g., GPS, camera) are
calibrated and the battery is sufficient for the flight.
QA/QC Check
- Confirm flight
plan covers the entire muckpile.
- Review sample images
post-flight to ensure clarity and no obstructions.
Field Image QA
1. Review sample
photos for blur & shadow.
2. Ensure full pile
coverage—no gaps on flight app grid.
3. Re‑fly missed
bands immediately.
Upload & Processing Steps
1. Create a Fragmentation project → Upload images
2. Set coordinate
system & identify scale objects.
3. Enable Fragmentation AI
and choose Custom size bins if needed.
4. Click Process
The AI performs photogrammetry, segmentation,
and PSD calculation automatically.
QC Checklist After Processing
|
Check
|
Pass Criteria
|
Fix if Fail
|
|
Point‑cloud
completeness
|
No gaps or
floating points.
|
Add images; re‑process
High‑Quality.
|
|
Segmentation
accuracy
|
≥ 90 %
correct outlines in spot check.
|
Use Edit
Rock Nodes; split/merge.
|
|
PSD curve shape
|
Smooth S‑curve;
logical D50/P80.
|
Verify scale
& GSD; adjust bins.
|
|
Histogram
outliers
|
< 5 %
in extreme bins (unless expected).
|
Inspect for
blur/shadow in images.
|
Editing & Customization
- Custom
bins: Settings → Size Ranges.
- Node
editing: Split oversized
polygons or merge over‑segmentation artefacts.
- Click Re‑run to refresh
stats.
Reporting & Interpretation
1. Open Analytics → Fragmentation Report
2. Review P80,
P50, Uniformity Index, Fines %.
3. Export PDF / CSV and
share.
4. Compare to blast
design targets; log variances.
Common Pitfalls & Quick Fixes
|
Symptom
|
Root Cause
|
Remedy
|
|
Excess fines
|
Too high
altitude / noise.
|
Fly 15–20 m;
fly slower
|
|
P80 spike
|
Add scale object
if possible
|
Use calibrated
bars next run.
|
|
Smeared rocks
|
Motion blur.
|
Shutter
≥ 1/1000 s; slower UAV speed.
|
|
Processing fail
|
< 60
images / low overlap.
|
Add images; re‑plan
grid.
|
Common
Issues
- Poor
Image Quality : Blurry or
shadowed images lead to inaccurate rock detection.
- Incomplete
Coverage : Missing sections of
the muckpile result in incomplete analysis.
- AI
Misidentification : Shadows,
debris, or complex rock shapes cause errors in boundary detection.
Solutions
- Image
Quality : Refly the muckpile
under better lighting or adjust camera settings for higher resolution.
- Coverage : Review flight logs to confirm full coverage and plan
additional flights if needed.
- AI
Errors : Use manual editing tools
to correct boundaries or contact Strayos support for advanced troubleshooting.