This module implements a highly extendable detection-tracking framework for robotic navigation. A basic detection pkg detectron2_detector and a tracking pkg kf_hungarian_tracker are provided.
A minimum representation of a dynamic obstacle includes position, velocity and bounding box. More features could be included to satisfy special need.
In detectron2_detector, Detectron2 is applied on RGB image to extract mask for target object and then RGBD image is used to estimate 3D position.
In kf_hungarian_tracker, Kalman filter is used to tracking dynamic obstacles in 3D space and Hungarian algorithm is used to associate obstacles and detection results.
- Ubuntu 20.04
- ROS2 foxy
- python3.8
