rjean/tracker

Real-Time Tracking Turret project using the Coral Edge TPU and mobile-net SSD.

Tracker

Small hobby project for an autonomous embedded device that can track things using a pan tilt system using the RaspberryPI.

The idea behind that is to be able to build a unofficial "Portal Sentry Turret: https://theportalwiki.com/wiki/Sentry_Turret".

There is a little bit of information in my blog about this project

Setup.

Install the required packages:

pip3 install -r requirements.txt

Make sure the gpio daemon is started when the raspberry pi boots:

sudo systemctl enable pigpiod

Architecture

Communication

A custom built "rtcom" library is used for communication.

It works by broadcasting endpoints on the network over UDP.

User Interface

The UI is handle by the video_reader.py file. It should be run on a PC, not on the raspberry pi.

As of now, it diplays the feed from the follower.

python video_reader.py

Camera

Make sure the camera is enabled:

sudo raspi-config
interfacing options->camera->enable the camera

AI

The controller requires a Coral USB Accelerator for embedded inference. https://coral.ai/docs/accelerator/get-started/.

curl -OL "https://github.com/google-coral/edgetpu/raw/master/test_data/ssd_mobilenet_v2_coco_quant_postprocess_edgetpu.tflite"

Run at startup

See https://www.tomshardware.com/how-to/run-script-at-boot-raspberry-pi

Edit crontab for the user:

pi@turret:~ $ crontab -e

And add that line:

@reboot bash /home/pi/follower/start.sh > /home/pi/follower/start.log 2>&1

Starting the controller.

On the raspberry pi, run

python3 controller.py