First steps in tracking drones for automatic control of swarms in high-density GPS denied environments
Current process:
-
get the following videos from the USRC drive and put them in \videos: a. flight1_nickphone.mp4 b. flight2_nickphone.mp4 c. flight1_saanviphone.MOV d. flight2_saanviphone.MOV
-
run videoslicer.py This guy pretty simply slices out each individual frame and gives them an index and a timestamp for a name.
-
run framesync.py I manually went into the 3k+ frames from the videos and found where they lined up (first frame with prop spin). Then this file takes every 10 frames from that point to what I decided was a good time to stop (landing). It puts them in \frames_synced
-
run MOG2_main.py It uses the MOG2 alg to separate out the pixels that are changing (moving) over time. Then it uses the cv2 findContours to make a bounding box around the biggest moving object. So yeah this would NOT work for a swarm AT ALL. It literally just finds the biggest moving blob lol.
- Theres a bunch of other files but they're like me messing around and iterating with chatgpt cuz I don't trust myself to manage my work in one file sometimes. U can probably igonre most of them. motiontracker.py is fun to look at but anything that mentions csrt does not work rn.