pedestrian-counting

a program to count pedestrians and bikers with a raspberry pi and OpenCV

TODO

• clean up code
• add kalman filter
• target class (might be deprecated)
• algorithms
• accurate counting
  • distance error?
  • discretization
  • overlap
• make sure everything is properly documented
  • use readthedocs.org instead of markdown
• add more options to the settings
• test live stream option
• get more test videos

counter.cpp

• main counting program

object.h and .cpp

• object class
• essentially a center point that id's an object being tracked

image_input.h and .cpp

• image input class
• wraps the raspicamera and video file options up so that the other programs don't care

image_output.h and .cpp

• image output class
• abstracts outputing to file or to the screen

trackers.h and .cpp

• static class that basically just a bunch of tracking algorithms
• matches object or targets from frame to frame

get_background.cpp

• simple program to average a video and find background
• not working that great, but could be due to shaky video

useful_functions.cpp

• a couple functions that I wanted to have available in multiple places

Also good to mention

• make sure you add the opencv stuff to your path: [path to opencv]/include
  • in our case /usr/local/packages/opencv/include
  • or do the -I thing when compiling
  • also -L for the libraries

Acknowledgements

Much of this code is based on helpful tutorials available here:

• http://docs.opencv.org/3.1.0/d5/d07/tutorial_multitracker.html#gsc.tab=0
• https://www.youtube.com/user/khounslow/featured
• http://docs.opencv.org/3.1.0/d1/dc5/tutorial_background_subtraction.html

Also:

• http://opencv.org/