AI Auto-Detection
Let AI automatically detect and trace paved area boundaries from satellite imagery.
How It Works
ProPaving uses computer vision AI to analyze satellite imagery and automatically identify paved surfaces. The system can detect:
- Parking lot boundaries
- Drive lanes and aisles
- Landscape islands (as cutouts)
- Building footprints to exclude
Subscription Required: AI auto-detection is a premium feature available to subscribers.
Using Auto-Detection
Position the Map
Navigate to the area you want to detect. Zoom in so the parking lot fills most of the screen (60-80% coverage is ideal).
Click Auto-Detect
Click the sparkle button in the toolbar. The button is disabled during drawing or when zoomed out too far.
Wait for Analysis
The system captures the current view and sends it for AI analysis. This typically takes 10-15 seconds. You'll see status messages: "Capturing imagery..." then "Analyzing paved area..."
Review Results
The detected boundary appears on the map with a results overlay showing confidence and options.
Understanding Detection Results
Confidence Score
Each detection includes a confidence percentage indicating how certain the AI is about the result:
Detection is likely accurate. Safe to accept directly.
Review the result. Consider refining before accepting.
Manual tracing may be more accurate. Consider rejecting.
Analysis Notes
The results overlay includes notes explaining what was detected, any issues found (shadows, unclear boundaries), and whether islands were identified.
After Detection
You have three options after the AI detects a boundary:
Accept
Convert the detection directly into a saved measurement. Use this when the boundary looks accurate.
Refine
Enter edit mode to adjust the boundary vertices before saving. Perfect for making small corrections.
Reject
Discard the detection and try again, or draw manually instead.
Refine Mode
When you click "Refine", you enter edit mode where you can adjust the boundary:
- Drag vertices: Move corner points to adjust the boundary
- Add vertices: Click on edge midpoints to add new points
- Remove vertices: Right-click on a vertex to delete it
- Live area: The square footage updates as you edit
When satisfied, click "Accept" to save your refined measurement, or "Cancel" to discard changes.
Tips for Best Results
Zoom Level Matters
Zoom level 18-19 works best. Too far = lot too small. Too close = can't see boundaries.
Center the Lot
Position the parking lot in the center of the viewport with some surrounding context.
Clear Imagery
Recent, high-resolution imagery with good contrast between pavement and landscaping works best.
Avoid Shadows
Building shadows can confuse detection. Try historical imagery if current view has shadows.
Simple Lots First
Detection works best on lots with clear boundaries. Complex shapes may need refinement.
Don't Force It
If detection consistently fails, manual tracing is faster than repeated attempts.
If Detection Fails
Sometimes detection may not work. Here's what to do:
"Viewport too small"
Zoom out slightly so more of the parking lot is visible, then try again.
"Analysis timed out"
The AI took too long. Try zooming to a different level or waiting a moment and retrying.
"Could not detect boundaries"
The AI couldn't identify a clear paved area. Try different imagery or use manual tracing.
"Rate limit exceeded"
You've made too many detection requests. Wait a minute before trying again.