pageauc/face-track-demo

Raspberry Pi Camera Pan/Tilt Face and Motion Tracking Demo

Uses python & Opencv to Track x,y position of moving object/face in camera view

Can use openelectrons pan/tilt assembly per http://www.mindsensors.com/rpi/33-pi-pan
or
gpiozero and pan/tilt servos connected to appropriate RPI gpio pins per https://github.com/RPi-Distro/python-gpiozero

Quick Install

Easy Install of face-track-demo onto a Raspberry Pi Computer with latest Raspbian Stretch (recommended).

for pi-pan openelectrons controller

curl -L https://raw.github.com/pageauc/face-track-demo/master/face-track-install-pipan.sh | bash

for gpiozero servo control

curl -L https://raw.github.com/pageauc/face-track-demo/master/face-track-install-gpiozero.sh | bash

From a computer logged into the RPI via ssh(Putty) session use mouse to highlight appropriate command above, right click, copy.
Then select ssh(Putty) window, mouse right click, paste. The command should download and execute the github script and install the face-track-demo project.
This install can also be done directly on an Internet connected Raspberry Pi via SSH (Putty), console or gui desktop terminal session and web a browser if you wish to copy the link.
Note - a Raspbian sudo apt-get update and upgrade will be performed as part of install so it may take some time if these are not up-to-date.

Manual Install

From logged in RPI SSH session or console terminal perform the following.

for pi-pan openelectrons controller

wget https://raw.github.com/pageauc/face-track-demo/master/face-track-install-pipan.sh
chmod +x face-track-install-pipan.sh
./face-track-install-pipan.sh

for gpiozero servo control

wget https://raw.github.com/pageauc/face-track-demo/master/face-track-install-gpiozero.sh
chmod +x face-track-install-gpiozero.sh
./face-track-install-gpiozero.sh

How to Run

OpenElectrons

The face-track-install.sh program will require an openelectrons pan/tilt assembly and servo controller installed and tested see http://www.mindsensors.com/rpi/33-pi-pan. See web site for installation details. This program installs and uses servo blaster and the pipan python library. See the face-track-install-pipan.sh for details. Support programs and utilities are installed in the /home/pi/pi-pan folder. These can be used to test the openelectrons pan/tilt operation. From SSH session, console or GUI desktop terminal session execute the following commands

cd ~/face-track-demo
./face-track-pipan.py

gpiozero

The face-track-install-gpiozero.sh will require pan/tilt servos connected to the appropriate RPI gpio pins per details and documentation links at https://github.com/RPi-Distro/python-gpiozero From SSH session, console or GUI desktop terminal session execute the following commands

cd ~/face-track-demo
./face-track-gpiozero.py

GUI or SSH display

Default is Jessie GUI desktop display. Use Nano to Edit config.py variables. default is window_on=True to display the opencv tracking window on GUI desktop.
See other variables and descriptions for additional variable customization settings.

The program can be run headless if required by setting config.py variable window_on=False

Face and Motion Track Demo - Basic concept of tracking moving objects

This Demo program detects motion and/or faces in the field of view and uses opencv to calculate the largest contour above a minimum size and return its x,y coordinate. The object is then tracked using the camera pan/tilt (within the range of motion available)

Motion Track Demo YouTube Video http://youtu.be/09JS7twPBsQ
GitHub Repo https://github.com/pageauc/face-track-demo

Introduction

I did quite a bit of searching on the internet, github, etc, but could not at the time find a similar python picamera implementation that returns x,y coordinates of the most dominate moving object in the frame although some came close. I have added face tracking.

If motion is detected then the largest moving contour will be tracked and a circle will indicate the center of the opencv contour
If no motion is detected for a while then face detection will be started.
If a face is found then it will be tracked and a rectangle will highlight the face contour.
If a face cannot be found after a specified number of retries then detection will revert back to looking for motion.
If there there is no face or motion detected for a longer delay then the camera will pan and tilt to look around until motion or a face is detected.

This setup may need to be tuned for your needs using the variables in the config.py file. I have tested this on a RPI 3 (quad core) and RPI B (single core). Performance on the RPI3 is reasonable although pretty laggy on the single core RPI B. Face detection takes a bit longer than regular opencv motion detection.
You may want to change the cascade file path to test other body parts per cascade files.
These can be found in the folder

/usr/share/opencv/haarcascades

Change the config.py file face_haar_path variable.

Prerequisites

IMPORTANT - You must have the RPI connected to a network with internet access during the installation. No internet connection is required otherwise.
For best results use a Quad core Raspberry Pi computer (recommended). You must be running with an up-to-date raspbian jessie distro and a RPI camera module installed on an OpenElectrons Pan/Tilt assembly that has been installed, configured and tested. Required python and servoblaster files and library Dependencies will be installed/upgraded per face-track-install.sh script depending on your previous installs. Note a sudo apt-get update and upgrade will be performed as part of the automated install so it may take some time depending on how up-to-date your system is.

Disable Camera LED

To disable the red LED you simply need to add the following line to your config.txt file :

disable_camera_led=1

To edit the config.txt file you can use Nano :

sudo nano /boot/config.txt

Trouble Shooting

if you get an opengl error first check that you are running the Desktop. If there is still a problem then see this article about installing opengl on a RPI P2 https://www.raspberrypi.org/blog/another-new-raspbian-release/

Otherwise install opengl support library per following command then reboot.

sudo apt-get install libgl1-mesa-dri

Edit the config.py file and set variable window_on = True so the opencv status windows can display camera motion images and a motion circle or face rectangle marking x,y coordinates as well as the threshold images. The circle diameter can be change using CIRCLE_SIZE variable.
You can set window_on = False if you need to run from SSH session. If debug=True then status information will be displayed without a GUI desktop session.

Credits

Some of this code is based on a YouTube tutorial by Kyle Hounslow using C here https://www.youtube.com/watch?v=X6rPdRZzgjg

Thanks to Adrian Rosebrock jrosebr1 at http://www.pyimagesearch.com for the PiVideoStream Class code available on github at https://github.com/jrosebr1/imutils/blob/master/imutils/video/pivideostream.py

---------- Other Raspberry Pi Projects Based on Motion Tracking ------------

speed-camera.py - Object (vehicle) speed camera based on motion tracking

Tracks vehicle speeds or other moving objects in real time and records image and logs data. Now improved using threading for video stream and clipping of area of interest for greater performance.

GitHub Repo https://github.com/pageauc/rpi-speed-camera
YouTube Speed Camera Video https://youtu.be/eRi50BbJUro
RPI forum post https://www.raspberrypi.org/forums/viewtopic.php?p=1004150#p1004150

cam-track.py - Tracks camera x y movements

Uses a clipped search image rectangle to search subsequent video stream images and returns the location. Can be used for tracking camera x y movements for stabilization, robotics, Etc.

GitHub Repo https://github.com/pageauc/rpi-cam-track
YouTube Cam-Track Video https://www.youtube.com/edit?video_id=yjA3UtwbD80
Code Walkthrough YouTube Video https://youtu.be/lkh3YbbNdYg
RPI Forum Post https://www.raspberrypi.org/forums/viewtopic.php?p=1027463#p1027463

hotspot-game.py - A simple motion tracking game

The game play involves using streaming video of body motion to get as many hits as possible inside shrinking boxes that randomly move around the screen. Position the camera so you can see body motions either close or standing. Pretty simple but I think kids would have fun with it and they just might take a look at the code to see how it works, change variables or game logic.

GitHub hotspot-game Repo https://github.com/pageauc/hotspot-game
YouTube Hotspot Gam Video https://youtu.be/xFl3lmbEO9Y
RPI Forum Post https://www.raspberrypi.org/forums/viewtopic.php?p=1026124#p1026124

----------------------------------------------------------------------------

Have Fun
Claude Pageau
YouTube Channel https://www.youtube.com/user/pageaucp
GitHub Repo https://github.com/pageauc