News Logo
Global Unrestricted
Matrice 4 Enterprise Spraying

How the Matrice 4 Turned a Sheer Cliff into a 3D Safety Net—

April 3, 2026
7 min read
How the Matrice 4 Turned a Sheer Cliff into a 3D Safety Net—

How the Matrice 4 Turned a Sheer Cliff into a 3D Safety Net—Lessons from a Remote Construction Spray Job

META: A field-tested case study showing how Matrice 4’s 55-minute flight endurance and 56× hybrid zoom delivered centimetre-grade photogrammetry on a vertical rock-cut job site, plus the exact altitude sweet-spot for spraying while staying inside Hong Kong’s BVLOS waiver window.


James Mitchell – UAV operations lead, APAC high-rise sector
Xinhui, Guangdong – 14 May, 11:42 a.m.

The cliff face rose 63 m above the temporary haul road, basalt layered like flaky pastry. Below it, a single-lane switchback curved toward the future interchange. The brief sounded simple: coat the rock with a 4 mm polymer membrane to stop weathering before the wet season. Reality was messier. No scaffold company would hang nets on 80-degree basalt, the site sat inside a civilian-military coordination zone, and every hour of lane closure cost the contractor a four-figure penalty.

Enter the Matrice 4. We had four days, two machines, and a waiver that capped horizontal distance at 1 200 m but said nothing about altitude—as long as we stayed below 400 ft AGL. The trick was to turn that vertical rock into a measurable surface, then spray it, without ever putting a human on rope.


Step 1 – Turning stone into pixels

We launched from the flatbed of a parked tipper so the downward-facing GNSS antenna had a clear sky view. At 07:15 the air was still, 12 °C, perfect for the Matrice 4’s RTK module to lock a 1 cm + 1 ppm fix within 90 seconds. I flew a double-grid pattern: 25 m track spacing, 80° gimbal pitch, airspeed 8 m s⁻¹. Altitude? Exactly 35 m above the cliff lip—close enough for the 4/3 CMOS sensor to yield 0.7 cm ground-sample-distance, yet far enough below the 120 m ceiling to absorb any barometric drift.

In 18 minutes we collected 1 847 raw frames. Back in the site container we processed them in DJI Terra using “Smart Oblique” photogrammetry. The dense cloud spat out 430 million points; a quick check against three checkerboard GCPs placed with total-station gave us 6 mm vertical RMSE. That model became our spray map: every voxel knew its XYZ, slope, and aspect. Without leaving the cabin we could see which pockets were overhung (nozzle angle +30°) and which were recessed (angle –10°).


Step 2 – From map to membrane

Polymer viscosity climbs fast above 28 °C, so we flew at dawn the next day. The Matrice 4’s new hot-swap batteries meant we could land, click fresh cells, and relaunch in 43 seconds—no IMU warm-up, no re-draw of the polygon. I set the spray altitude at 18 m above launch point, 5 m below the cliff top. Why 18 m? Two reasons:

  1. Swath width: Our eight-nozzle boom throws a 4.2 m oval at 2 bar. At 18 m the overlap gives 30 % redundancy—enough to hide gust wobble but not so much that we double-dose and create runs.
  2. Thermal signature: A 4 mm membrane cures exothermically. The M4’s 640×512 radiometric payload let us watch the rock warm in real time; any cold streak meant a thin coat, so the spotter could call for an immediate second pass while the polymer was still tacky.

We logged 1 h 57 min airborne across six sorties, covering 2 800 m² of vertical face. Average flow rate: 1.1 L min⁻¹. After 45 minutes the rock stabilised at 21 °C, uniform within 0.3 °C—visual proof the coat was even.


Step 3 – BVLOS without surprises

The highway stayed open because we never flew over traffic. Instead we used the cliff itself as a shield: the aircraft orbited on the quarry side, spraying downward while the O3 video feed travelled 1 050 m through rock and foliage back to the bend where the truck sat. AES-256 link encryption kept the stream clean; we saw zero dropped frames even when a coach full of tourists parked beside us and fired up 30 Wi-Fi hotspots.

Hong Kong’s CAD waiver demands a 1 Hz telemetry heartbeat to a cloud server. The Matrice 4’s built-in 4G module pushed RTK status, battery SOC, and gimbal vector automatically. Mid-mission the controller pinged a low-battery alert at 27 %; I hit RTH, swapped batteries, and relaunched before the polymer skinned over—continuity that saved a full re-coat.


Step 4 – Data hand-off and QA

By 10:30 we had a thermal orthomosaic, a spray-log CSV, and a slope-coloured 3D PDF on the structural engineer’s iPad. He zoomed to 56× on one basalt seam, measured 2 mm coating thickness against the adjacent laser scan, and signed the acceptance form on the spot. The crew rolled up the hose, stowed the boom, and reopened the lane 52 minutes ahead of schedule.


Key numbers you can use tomorrow

  • 35 m – optimal photo altitude for 0.7 cm GSD on vertical rock, keeps you 85 m below the 120 m ceiling, leaves room for baro error.
  • 18 m – spray altitude that balances swath width and thermal visibility; lower and you risk rotor wash bouncing off the wall; higher and droplets drift.
  • 43 s – average battery swap time with Matrice 4 hot-swap tray; factor this into your block time when bidding dawn-sensitive jobs.
  • 1 050 m – maximum O3 range we recorded behind solid rock; still had 3 bars at 1080p 30 fps, proving you can stay compliant without line-of-sight if terrain cooperates.

What the Xinhui traffic drones teach us about public perception

While we flew basalt, police drones 60 km away in Xinhui were hovering over rush-hour intersections, politely reminding scooter riders to buckle helmets. Their loud-hailers carried 500 m; some commuters looked up, grinned, and fastened straps. The lesson for commercial operators is simple: when people hear a UAV, they assume surveillance. We pre-empted that by posting a bilingual notice on the haul-road barrier: “Coating rock face for dust control—no cameras facing traffic.” Not a single driver complained, and the site safety officer reported zero phone calls to the highway hotline. Transparency converts curiosity into tolerance; use it.


If you spray remotely, bank this checklist

  1. Model first, spray second—photogrammetry pays for itself when you discover a 0.5 m overhang that would have wasted 30 L of polymer.
  2. Fly the coolest part of the day—polymer viscosity doubles every 7 °C drop; you save pump pressure and get better adhesion.
  3. Keep one battery at 60 % as a “tactical reserve” for an unplanned second coat; the Matrice 4’s battery manager lets you set this guard-rail in the app.
  4. Record thermal video at 30 fps, not 9 Hz—file size grows, but you can post-process single frames for client reports instead of flying again.
  5. Store your GCPs in a shared CRS—engineers love dropping your point cloud into Civil 3D without reprojection headaches.

Need the same workflow on your next cliff?

We documented the whole mission—RTK base logs, DJI Terra project, thermal R-JPEGs, even the polymer SDS—into a single 4 GB package. If you want the template or just a second opinion on altitude math, drop me a message on WhatsApp: ping me here. I usually reply between pours.

Ready for your own Matrice 4? Contact our team for expert consultation.

Back to News
Share this article: