Object Detection with Multiple Lidar Fusion

The aim of this project is to concatenate two point cloud with rigid body transform and detect objects by using PCL (Point Cloud Library). The project includes;

• Taking two different point clouds with different axises and transform one of them in order to make their z axis the same. The algorithm publishes the transformed point cloud data in /transformed topic.

  rosrun lidar_fusion transform_pointcloud_main

• Taking the transformed point cloud for concatenating it with the second point cloud. The algorithm publishes fused point cloud data in /fused topic.

  rosrun lidar_fusion fusion_main

• Taking the fused point cloud in order to detect objects. The algorithm publishes found objects poses in /ObjectDetection/poses , clustered point clouds in /ObjectDetection/cloud_filtered and markers of the objects in /ObjectDetection/markers topics.

  rosrun lidar_fusion detection_main

Parameters

transform_pointcloud

Parameters	Definition
first_lidar_topic	The first point cloud topic.
second_lidar_topic	The second point cloud topic. This point cloud gets transformed but if required both point clouds can get transformed.
lidar_pos_x	Desired value of transformation on x axis.
lidar_pos_y	Desired value of transformation on y axis.
lidar_pos_z	Desired value of transformation on z axis.
lidar_rotation	Desired value of transformation on yaw.

fusion

Parameters	Definition
first_pc	The first point cloud topic.
second_pc	The second point cloud topic.

detection

Parameters	Definition
sensor_model	Sensor model. (In this project Velodyne HDL-323 is used.)
z_axis_min	Desired value of minimum z axis of region of interest for removing the ground.
z_axis_max	Desired value of maximum z axis of region of interest for removing the ceiling.
cluster_size_min	The minimum number of points that a cluster needs to contain in order to be considered valid.
cluster_size_max	The maximum number of points that a cluster needs to contain in order to be considered valid.
lidar_topic	The input point cloud topic.

Install

This package has three main dependencies:

sudo apt install ros-$ROS_DISTRO-velodyne-simulator

sudo apt install ros-$ROS_DISTRO-perception-pcl

sudo apt install ros-$ROS_DISTRO-pcl-conversions

ROSDep will take care of the other dependencies.

rosdep install --from-paths src --ignore-src -r -y

You can run via roslaunch lidar_fusion overall.launch