ROS2 3D bouding box object detection from pointclouds

This is a package for a baseline 3D bounding box detection and tracking using pointcloud data.

It’s based on this package https://github.com/praveen-palanisamy/multiple-object-tracking-lidar

Design and algorithm

1. Detect from clustering Unsupervised euclidean cluster extraction
2. Track tracking (object ID & data association) with an ensemble of Kalman Filters
3. Classify static and dynamic object

Installation

colcon build

Usage

ros2 run pcl_object_detection pcl_object_detection --ros-args --param-file config/config.yaml

Topics

Subscriptions:

• Name: /filtered_clouds Type: sensor_msgs::msg::PointCloud2

Publishers:

• Name: viz Type:visualization_msgs::msg::MarkerArray

• Name: cluster Type: sensor_msgs::msg::PointCloud2

• Name: obj_id Type: std_msgs::msg::Int32MultiArray