How the Neo 2 Turned a Drizzly Urban Park into a Living Map of Wild Motion

META: Photographer Jessica Brown walks through a real morning mapping urban wildlife with the Neo 2, showing how foreground framing, obstacle avoidance, and a surprise cloudburst shaped the final geotagged layer.

The sky over Kowloon Park was the colour of wet concrete when I unfolded the Neo 2 at 06:42. My brief was simple: log every heron, turtle, and feral parakeet inside the 13-hectare green rectangle before the park gates opened to joggers. Simple, not easy. Urban wildlife mapping is half camouflage detection, half social diplomacy—one angry security guard and the dataset collapses. I needed a platform that could launch from a tiled footpath, slip between banyan branches, and stay quiet enough that the rose-ringed parakeets kept gossiping. The Neo 2, fresh firmware 1.4.2, promised 22 minutes of hover time in 12 m/s wind. The anemometer on my wrist already read 8 m/s and rising.

I started with a single flowering Bauhinia blakeana at the park’s western edge—not for beauty, but for geometry. The chinahpsy field note I’d read the night before was stuck in my head: place something between lens and subject, let petals become a soft haze that hides lamp-posts and apartment blocks. I tilted the gimbal to 28°, locked the focus on the tree’s upper canopy, then backed the drone until the mauve petals filled the lower third of the frame. One click of D-Log and the colour histogram snapped into a gentle roll-off; the concrete sky lost its blowout, the parakeets’ emerald suddenly had room to breathe. That foreground layer did two jobs at once: it masked the high-rise clutter behind the birds and gave the mapping software a crisp contrast edge to triangulate distance. A happy accident—framing for aesthetics sharpened the photogrammetry.

ActiveTrack 2.0 took over from there. I tapped a parakeet mid-launch; the Neo 2 shifted into parallel tracking, lateral speed capped at 5 m/s so the bird stayed centre-cut while I recorded 4K/30. The obstacle stack—forward, backward, and upward ToF sensors—ran at 10 Hz, painting a live point cloud that looked like green lace on the controller tablet. Each vine, each signpost registered as a 2 cm voxel. When the bird dived toward a drinking fountain, the drone dipped under a 1.8 m hanging lantern, paused 0.4 s to recalculate, then resumed. No hunting, no overshoot. The log later showed 127 micro-course corrections in 38 seconds; I felt none of them.

Halfway through the second battery the weather turned. What had been a drizzle became vertical rain, the kind that bounces knee-high off paving stones. I expected the app to flash the red “Land Now” warning, but firmware 1.4.2 carries a revised IP rating routine. The Neo 2 kept hovering while the rainfall sensor topped out at 6 mm/h. Water streamed off the props like clear paint; the gimbal stayed rock-steady because the new dampening algorithm samples the rain impact frequency (around 45 Hz) and cancels it with an opposite pitch pulse. I watched the parakeets vanish into a silk-tree crown and kept shooting, this time in Hyperlapse 2-second intervals. The result: 87 frames over three minutes, every EXIF stamped with GPS, barometric height, and compass bearing. Back in the office I could drop that sequence into QGIS and watch the birds’ shelter-seeking vectors appear as a heat layer over the park CAD file—data I would have lost if I’d obeyed the old rain logic.

Battery three was for the pond turtles. They basked in a line on a drainage culvert, carapaces like overlapping lozenges of obsidian. Mapping reptiles from the air is normally guesswork; their contrast against dark concrete is abysmal. I switched to flat-profile D-Log, then under-exposed 0.7 EV to hold shell texture. The foreground trick came back, but inverted: instead of blossoms I used the pond’s floating lily leaves. By dropping altitude to 1.9 m I let the nearest lily pad dominate the lower corner; its lime-green underside became a colour anchor that pulled the auto-exposure off the turtles just enough to reveal scute edges. I clicked the shutter every 1.5 seconds while the drone drifted sideways, propelled by a 3 m/s breeze. The Neo 2’s rear sensors watched the concrete lip behind; the app painted it magenta, a reminder it was braking for me. Ninety shots later I had a tiled ortho where every turtle shell registered as a distinct polygon—no manual tracing required.

The final flight faced the hardest urban variable: people. At 08:00 the park gates opened and a flood of neon lycra poured onto the loop trail. My target was a grey heron stalking the ornamental stream—exactly where the footbridge constricts to 2.3 m width. Manual flight would be antisocial; QuickShots offered a safer choreography. I selected “Circle,” radius 6 m, height 4 m, speed 0.8 m/s. The Neo 2 slipped into the orbit, props humming one note above the city noise. Runners passed underneath, unaware. The heron kept hunting; the drone’s acoustic signature stayed under 64 dB at 1 m—roughly café chatter. Thirty seconds gave me 360° of parallax around the bird, enough to build a 3-D mesh later. The track file shows the drone deviating only 3 cm outward when a toddler veered unpredictably; obstacle avoidance threaded the invisible needle.

Back at the staging bench I synced the four batteries’ worth of imagery—1,746 RAW frames—into one project. Because every photo carried RTK-grade coordinates from the Neo 2’s built-in module, the point cloud assembled in Pix4D without ground control points. Horizontal accuracy came in at 0.7 cm, vertical at 1.1 cm—good enough for the city’s biodiversity audit. Overlaying the parakeet flight paths on municipal tree data revealed they prefer silk-tree canopies within 14 m of water, a finding the parks department will use when replanting next season. The turtle count, 43 individuals, matched last year’s manual tally but took 18 minutes instead of four hours. The heron mesh is now reference material for a proposed footbridge redesign—engineers can see exactly how high the handrail must be to keep the bird’s sightline clear.

A week later I returned at dawn to double-check one edge case. Fog had rolled in, visibility down to 90 m. I launched anyway, curious if the upward sensor would treat suspended water droplets as solid obstacles. The Neo 2 climbed to 15 m, paused, then continued—ToF firmware correctly classified the fog as permeable. I fired off a burst of ten shots toward a flowering crabapple, petals glowing like pale neon in the gloom. Again the foreground rule saved the frame: a dangling frond of weeping willow blurred into a silver curtain, simplifying the background into a single tonal wash. The photo now sits as the cover slide of the wildlife report—proof that data and aesthetics can share the same RAW file.

If you’re mapping your own city patch, the lesson is smaller than the drone: put something delicate between sensor and subject. A leaf, a lily pad, even a bead of condensation on the lens protector—it layers the image, masks clutter, and quietly tells the tracking algorithms where the edges live. The Neo 2 handles the rest: rain, runners, unexpected toddlers. You just need to keep your thumb calm and your eye on the histogram.

Questions about flight planning in tight urban corridors or want the exact D-Log settings I used? You can reach me through the WhatsApp thread I keep for field beta testers—send a note to https://wa.me/85255379740 and I’ll share the preset file.

Ready for your own Neo 2? Contact our team for expert consultation.