Yard Mapping Project

Drone mission planner + workflow for mapping a yard with a DJI Mavic Air 2 via Litchi.

Files

• yard_mission_planner.py — the planner script
• yard_polygon.csv — yard polygon vertices (lat,lon per row). Edit this with your actual corners.
• example_yard_polygon.csv — sample polygon for reference
• mission_config.yaml — mission parameters (annotated; all settings have script defaults)
• mission_output/ — generated CSVs + KML preview

Setup

pip install shapely simplekml pyyaml

Usage

# Edit yard_polygon.csv with your actual yard corners first
python yard_mission_planner.py

Outputs land in ./mission_output/:

• mission_nadir.csv — Litchi import for the top-down grid mission
• mission_orbit.csv — Litchi import for the perimeter oblique orbit
• mission_combined.csv — both missions concatenated, single import for one continuous flight
• missions_preview.kml — open in Google Earth to verify before flying

Getting yard coordinates

Open Google Maps satellite view, right-click each corner of your yard, click the lat/lon at the top of the menu (auto-copies). Walk the perimeter in one direction (CW or CCW both fine) and paste each pair as a row in yard_polygon.csv.

Pre-flight: camera setup (DJI Fly)

Litchi's camera control surface for the Mavic Air 2 is limited — set these in DJI Fly first, then switch to Litchi. Settings persist on the drone.

• Photo mode: Single Shot (not AEB, HDR, Burst, Hyperlight)
• Format: JPG only (JPG+RAW writes are too slow and cause dropped photos)

Format the SD card in DJI Fly before flying (exFAT, U3/V30 or better). Copy off any photos you want to keep first — format wipes the card.

Pre-flight: Litchi global mission settings

This is the #1 source of grief. The Litchi CSV format only encodes per-waypoint values. Global mission settings (Path Mode, Heading Mode, Cruising Speed, etc.) are not stored in the CSV — they live in the Litchi app/Mission Hub and will silently override per-waypoint values if misconfigured.

The symptom of getting this wrong: drone flies through waypoints without stopping, and photos either don't trigger or only the first one fires.

After importing the CSV into Litchi, set these globals on the mission:

Setting	Value	Why
Path Mode	Straight Lines	Without this, the drone smooths through waypoints and never "arrives" — Take Photo actions never fire
Default Curve Size	0%	Forces a full stop at every waypoint; overrides per-wp curvesize if non-zero
Heading Mode	Custom	Uses per-waypoint heading values from the CSV (not POI, not Auto)
Default Gimbal Pitch Mode	Interpolate	Smoothly transitions gimbal between nadir (−90°) and orbit (−58°) sections
Rotation Directions	Managed	Picks shortest rotation per turn
Photo Capture Interval	disabled	We use per-waypoint Take Photo actions; intervals would double-shoot
Cruising Speed	3 m/s	Low enough that the drone can brake to a full stop at each waypoint
Finish Action	RTH	Drone returns home when mission ends (instead of hovering on the perimeter waiting for input)

Pre-flight: drone & app hygiene

• Re-import the CSV every time you regenerate it. Litchi snapshots the mission on import — saving a new file on disk does not update what Litchi will fly. Delete the old mission from Mission Hub, re-import the new CSV.
• Force-close DJI Fly before opening Litchi. Background DJI Fly intercepts the camera and breaks Litchi's photo triggers.
• Power-cycle the drone and controller between flights — clears wedged camera/gimbal state from the previous run.
• Verify camera triggers on the first 2–3 waypoints before letting the mission run unattended. Look for the photo counter incrementing or the shutter feedback (brief screen flash / audible click).

Pre-flight: altitude vs obstacles

The mission altitudes come from mission_config.yaml. Standard rule of thumb is 5–10 m clearance above the tallest obstacle (trees, wires, structures). GPS altitude drift is ±1–2 m and wind can push the drone down further, so a 2–3 m margin is too thin.

The KML preview (missions_preview.kml, open in Google Earth) is your final sanity check before flying.

Post-flight: process photos

1. Pull the SD card. Photo count should roughly match the waypoint count (planner prints both nadir and orbit waypoint counts on each run).
2. Spot-check ~10 photos at full res for sharpness; verify EXIF includes GPS coordinates (right-click → Get Info on Mac).
3. Process with WebODM (free, runs locally via Docker): drop both nadir and orbit photo sets into one task → orthomosaic + 3D model + point cloud.
4. View / analyze the orthomosaic GeoTIFF in QGIS (free).

Mavic Air 2 specifics

The Mavic Air 2 / Air 2S use Virtual Stick Control in Litchi — the mission isn't stored on the drone, it's driven moment-to-moment from the controller via DJI's MSDK 5. This makes timing-sensitive operations (waypoint stops, photo triggers) inherently less reliable than older drones (Mavic 2 Pro, Phantom 4) where missions are stored onboard.

Expect occasional photo drops even with everything configured correctly. If photos drop badly despite the checklist above, the fallback is DroneLink (free tier supports Air 2 with native waypoint missions and a different SDK path).