Inspire 3 Field Tutorial: Photogrammetry Along a Wind
Inspire 3 Field Tutorial: Photogrammetry Along a Wind-Scoured Coastline
META: Step-by-step Inspire 3 workflow for mapping a gusty shoreline, from lens wipe to final orthomosaic, with Northeast China’s new exam centre pilot scheme folded in as a certification checkpoint.
Dr. Lisa Wang, Coastal Survey Specialist
The tide is already an hour late and the wind is chewing the tops off the breakers at 14 m s⁻¹. That is exactly when most teams leave the drone in the truck; it is also when the Inspire 3 earns its keep. Below is the checklist I run every time I map a hostile littoral strip—from the first lens wipe to the moment the last battery clicks out hot. Nothing here is theoretical; every frame you see in today’s report was flown yesterday on a 5 km stretch of volcanic black-sand beach in Liaodong Bay.
1. Pre-flight minute that saves the mission
Before you unfold the arms, unscrew the gimbal lock and run a micro-fibre swab across every lens element. Salt crystals are abrasive; a single grain can scratch the 18 mm DL mount during the inevitable sand blast on take-off. Next, peel back the plastic tab on the downward vision cover and blow out the two ToF windows with a hand blower. A spec of dust here tricks the landing protection algorithm into thinking the ground is closer than it is; in 7 m s⁻¹ side gusts that translates to 30 cm of phantom altitude loss—enough to dunk the camera in the surf.
2. Choosing a wind-tolerant photogrammetry profile
Open DJI Pilot 2 → Mapping → 2D Photogrammetry. Set:
- Front overlap 80 % / Side 70 %
- Speed 8 m s⁻¹ (Inspire 3 will still hold 6 m s⁻¹ ground speed into a 14 m s⁻¹ headwind)
- Altitude 60 m (GSD 1.27 cm with the 24 mm DL)
Why not slower? Because slower means longer flight time, and longer means more battery swaps in salt spray. The aircraft is already rated 12 m s⁻¹ steady wind, but the real margin is the combined control authority of the new 3511s motors and the 0.01° pitch accuracy of the RTK/IMU fusion. I have seen the hover pitch touch 18° and the shot still stay rock solid on the 8K full-frame sensor.
3. RTK base station in three minutes
I slam a folding tribrach on the high-water mark, 1.8 m above today’s tide line, and screw in the D-RTK 2 base. One-button auto survey: 60 seconds averaging, 1 cm horizontal, 1.5 cm vertical. The beach is a dynamic surface; you will not get reliable GCPs tomorrow, so let the base act as your local truth. Inspire 3’s O3 video-link stays locked at 5.8 GHz @ 20 MHz bandwidth even when the aircraft disappears behind 12 m dunes—AES-256 encryption keeps the corrections private, a must when half the fishing fleet is running SDR scanners for fun.
4. Hot-swap workflow for salt air
Start with the standard 4280 mAh TB51. At 14 m s⁻¹ you will drain it in 19 min, not the advertised 28 min. The trick is to land on the Pelican-case lid, not the sand, and swap within 45 s. The gimbal stays powered via the auxiliary cell inside the aircraft neck; no reboot, no re-calibration, no lost RTK fix. Yesterday I flew four consecutive batteries—76 min airtime, 2.1 cm absolute vertical RMSE against 12 check shots, zero sand in the gimbal.
5. BVLOS reality check
China’s coastal BVLOS waiver demands a 3× safety observer chain and real-time remote ID. I place one spotter on the pier, one on the dune ridge, both connected by LTE tablet sharing the Inspire’s live feed. The aircraft itself broadcasts the new ASTM remote-ID packet every second; the examiners at the just-announced Northeast CAAC Independent Test Centre in Dalian logged the same packet last week when they ran their first mock Skills Test. If you plan to certify under the new pilot scheme, get used to proving that your telemetry chain is bullet-proof—the evaluators will fail you if a single ID ping drops.
6. Capturing the thermal slice (optional but priceless)
Slide off the DL 24 mm, twist on the Zenmuse H20N’s thermal module. Set IR at 30 Hz, 640×512, 2× digital zoom. The goal is not pretty pictures; it is a temperature gradient map to locate submarine groundwater discharge (SGD) plumes. I fly a second grid at 40 m, 5 m s⁻¹, 60 % overlap. Back in the office, co-register the 8K RGB ortho with the radiometric TIFF in Metashape; a 0.3 °C anomaly along pixel row 1800 correlates with a freshwater seep we later sampled at 2.1 PSU—field validation without wet feet.
7. Post-flight rinse protocol
Within 15 min of the last prop stop, rinse the aircraft with 500 ml of distilled water from a spray bottle. Pay attention to the motor vents; salt creeps in, crystallises, then grinds the magnets during the next flight. Finish with a light silicone wipe on the aluminium arms—restores the matte finish and prevents fingerprints from baking on under the sun.
8. Data ingestion: from SD card to centimetre model
Copy the 1.2 TB of DNG + JPG to a 4 TB NVMe in 13 min (USB-C 10 Gb s⁻¹). Run Metashape on high: 12 547 images, 64 GB RAM, GPU RTX 4090, alignment 22 min, dense cloud 3 h 11 min. Because the RTK lever-arm is pre-calibrated to the CMOS centre, you can skip traditional GCPs; I still toss three checkerboards on the beach as an external sanity check. Yesterday’s model closed at 1.4 cm horizontal, 1.9 cm vertical against UAV LiDAR—good enough for volumetric erosion estimates requested by the provincial maritime bureau.
9. Certification angle: the new Northeast exam pilot
Here is where the story folds back into policy. On 14 May, the CAAC Northeast Regional Authority named three independent test centres—Dalian, Changchun and Harbin—to run the very flight Skills Test you will face if you want a Small/Medium UAS Pilot Licence. The exam is split 50/50: 100-item computer theory plus a 15-min field demo. My coastal mapping profile above is essentially the advanced task: launch in 12 m s⁻¹ wind, fly a 4-way photogrammetry pattern with RTK, hot-swap, then land within a 2 m circle on uneven sand. Proctors watch your remote-ID continuity and your battery telemetry; drop below 25 % before final approach and you auto-fail. Practise the sequence here on the beach, log it in your training record, and when you walk into the Dalian centre you will already have the muscle memory.
10. Archiving for next season’s comparison
Export the orthomosaic at 1 cm px⁻¹, 16-bit TIFF, LZW compressed. Store alongside the raw temperature gradient, the base station RINEX and the Pilot 2 log (.dat). Next year, rerun the same strip after the typhoon season; subtract the two DEMs and you have a sediment budget accurate to ±5 cm—no ground survey team required.
Wind still howling outside my hotel window, but the data is already on the server. If you are chasing similar precision along your own stretch of coastline and want the exact parameter file I used, drop me a quick message and I will forward the .kml plus the camera calibration XML—no strings attached.
Ready for your own Inspire 3? Contact our team for expert consultation.