analyze-uav-detections

name	analyze-uav-detections
description	Analyze UAV detections produced by the YOLO pipeline.
homepage	https://github.com/alebalzan/luckfox_UAV_detection
metadata	{"nanobot":{"emoji":"🛸","requires":{"bins":["/root/luckfox_pico_rtsp_yolov5_UAV_demo/utilities/read_detections"]}}}

Analyze UAV Detections

Analyze UAV detections produced by the YOLO pipeline.

Steps

1. Dump the motion-feature JSON to a temporary file:

/root/luckfox_pico_rtsp_yolov5_UAV_demo/utilities/read_detections -j /tmp/detections.json

2. Read the file:

cat /tmp/detections.json

3. Analyze the JSON array. Each object is one decimated time window for one target:

timestamp_us     — end-of-window timestamp (microseconds since epoch)
trajectory.centroid  — [nx, ny] normalized position in frame [0..1]
trajectory.velocity  — [vx, vy] smoothed velocity in pixels/second
trajectory.speed     — smoothed scalar speed in pixels/second
trajectory.heading_rad — direction of motion in radians (atan2)
trajectory.samples   — number of raw detections averaged in this window
confidence       — mean detection confidence in this window
target_id        — YOLO target number

4. Report:
   • Current position and movement direction of each target
   • Speed trend (accelerating / decelerating / steady)
   • Approach angle relative to frame centre
   • Any targets that disappeared or reappeared

Notes

• The file /tmp/detections.json is on the device's RAM tmpfs; it is lost on reboot.
• Detection logs are stored persistently under /userdata/uav_detections/.
• If no detections are found, check that the main pipeline binary is running:

  ps | grep luckfox_pico
  

• To get raw bounding boxes instead of motion features, use --tail 20 for a table or -c /tmp/export.csv for full CSV.