You'll need OpenCV, Python (I use 2.7), and a webcam.
If you've got all three of those, it should be good to go.
Please tell me if it's not.
The file you want is [ motiondetection.py ].
IMPORTANT: While it's running, just hit the 'q' key to quit. Similarly, press 'm' to toggle motion detection.
(this should run straight out of the box.)
This tidbit of Python is (hopefully) representative of my second foray into OpenCV motion tracking.
I intend to develop something that can accurately discern humanoid motion (and will not raise false alarms for things like quick changes in lighting conditions).
Ideally, it'll be able to take a feed from a webcam in real time, highlight detected motion, and raise some sort of alarm if the motion detected appears to be humanoid - highlighting also the suspect motion in a screencap (or by emphasizing it in a video feed).
So: this will probably start with a simple import cv and we'll see where it goes from there.
While I know this is probably a mixture of reusing and reinventing the wheel, I feel like it is a good exercise (and will also be contributing to a personal project).
...the spec for which is quite interesting.
Did you know: True + True = 2?
- Reimplement with a nicer GUI
- maybe get the bounding boxes to be "prettier"?
- and/or more generally surround the motion
- also, notify the user of
- the 'mode' they're in
- uptime of the script
- number of motion events detected since startup
- Improve motion detection
- slow-ish or small movements in-frame are hard to detect
- would we be able to confuse this with camouflage?
- slow-ish or small movements in-frame are hard to detect
- Implement 'alerts'?
- probably takes place after the nicer gui implementation
I started with these very helpful copy-paste-and-play tidbits:
https://gist.github.com/burnto/1266515 http://stackoverflow.com/questions/3374828/how-do-i-track-motion-using-opencv-in-python
Yup.