Tracking camera

Overview

Tracking camera : This package controls a 3 axis 3D printed gimbal using computer vision to track human faces and/or body. The gimbal can also be controlled manually via an rqt interface. Everything works in a ROS environnement. In this package, you can find all the mechanical parts needed to build our project as well as all the code to make it work.

Install

Mechanical assembly

To assemble the mechanical parts you'll have to 3D print a few parts and buy some others. For complete instruction about how to procede, follow this guide

OpenCR setup

To control the motors with ROS you’ll need to put the OpenCR board into a usb to dynamixel bridge.

First, change your port permissions by typing this command line:

[user@machine ~]$ sudo usermod -a -G dialout <user>

Note that you have to logout/login to make the change available . Then follow the instructions here to make your board visible on Arduino IDE. You only have to follow the steps from 4. 1. Install on Linux to 4. 1. 5. 3. Port Setting

Now you can upload the following code File -> Examples -> OpenCR -> 10.Etc -> usb_to_dxl

Software

For this project, you will need a computer or Jetson Xavier as well as an OpenCR. First of all, you need to have Ubuntu 18.04 installed on your computer as well as ROS Melodic. Here are the installation steps :

http://wiki.ros.org/melodic/Installation/Ubuntu

Dependencies:

1. darknet_ros
2. dynamixel_workbench
3. usb_cam

sudo apt-get install ros-melodic-dynamixel-workbench
sudo apt-get install ros-melodic-usb-cam

darknet_ros is not yet supported on Melodic, so you will need to build it from the source.

https://github.com/leggedrobotics/darknet_ros

After cloning our repository, build in your catkin workspace :

git clone https://github.com/shovington/Tracking_camera.git
catkin_make

Usage

Once everything is installed, run our project with

roslaunch tracking_camera full.launch

To run only manual mode

roslaunch tracking_camera manual.launch

Here's what our user interface looks like when launched.

Config files

Darknet config

• darknet.yaml

Contains the camera configuration for darknet and which image topic to subscribe to.

Motors config

• config_motors.yaml

Contains the motors configuration : IDs, names, etc.

Yolo config

• yolov3.yaml

Contains yolo configuration : Accuracy threshold, classes to track.

Nodes

tracking_cam_plugin node

This is the only node in the package. It is a rqt plugin. You can control the gimbal manually using the UI sliders. It also subscribes to darknet and controls the gimbal to track faces when it is put in auto mode.

Subscribed Topics

• /dynamixel_workbench/dynamixel_state ([dynamixel_workbench_msgs/DynamixelStateList])

  Motors live states

• /darknet_ros/bounding_boxes ([sensor_msgs/NatSatFix])

  Position and dimension of detected objects

Run automated tests

Enter this line in terminal:

rostest tracking_camera test_test.test