January 14, 2026
Planespotters, start your drama
Project SkyWatch (a.k.a. Wescam at Home)
DIY sky cam tracks jets like a cop chopper—debate: bigger motors or more AI
TLDR: A DIY rig turns a cheap conference camera into a jet-tracking setup using prediction, stabilization, and flight data. Commenters cheered, then battled over AI detection vs the existing tracker and calls for real gimbal hardware, while giggling at the name “WESCAM.”
Internet tinkerers are losing it over a homebrew “sky turret” that turns a cheap pan-tilt-zoom church cam into a jet-stalking machine. The creator fuses visual tracking (CSRT, a fancy way to lock onto the plane’s texture), prediction with a Kalman filter (aka “where it’ll be in 200ms”), and a PID control loop (a smart way to smooth and steady the aim) with a digital “virtual gimbal” to kill jitter. It even cross-checks flights with ADS-B data so the overlay can say exactly what you’re watching. The crowd’s verdict: hype, questions, and delicious drama.
One camp is in pure awe—“rock solid” tracking, like a police helicopter but from your backyard. Another camp immediately shouts hardware upgrade: ditch the plastic gears, go direct drive with real gimbal motors, and watch the precision soar. The AI crowd pops in with “why not YOLO?”—suggesting object detection magic to auto-find planes instead of manual lock-on. Meanwhile, one joker side-eyed the name WESCAM (“wild brand for something that looks like a webcam’s secret agent cousin”), and the creator himself showed up, thrilled and hinting at adding thermal vision next.
It’s the perfect nerd soap opera: software vs hardware, prediction vs detection, and a budget rig that’s somehow serving military-grade vibes. SkyWatch: part backyard fun, part conspiracy joke, fully comment-section war zone.
Key Points
- •The project uses a consumer PTZ camera to emulate professional EO/IR gimbal tracking for aircraft.
- •Visual tracking is performed with OpenCV’s CSRT tracker to compute pixel error and maintain lock on target features.
- •A Kalman Filter smooths noise and predicts target state ~200 ms ahead, enabling feed-forward motor control.
- •A PID controller combined with feed-forward and dynamic speed limiting achieves responsive yet precise motion.
- •A digital stabilization layer (virtual gimbal) crops and shifts frames to counter mechanical jitter; ADS-B data provides identification context.