obstacledet's Introduction

Obstacle Detection Functionality

This project provides obstacle detection functionality based on the fusion of laser scanner and point cloud data, suitable for scenarios where a cleaning robot needs to navigate around obstacles.

Usage

Compilation

mkdir -p catkin_ws/src && cd catkin_ws/src
git clone https://github.com/SheldonFung98/ObstacleDet
cd ..
catkin_make
source devel/setup.sh

Launching

roslaunch obstacle_det detect.launch

Code Principle

Wall Detection Functionality

Method 1:

Crop the point cloud ROI, downsample the point cloud within the ROI, and grid the point cloud.
Use RANSAC to find a plane with normals pointing upwards, which serves as the ground.
Elevate this plane (to ensure all ground points are filtered out) and filter out points below the plane.
Project non-ground points onto the plane and publish the point cloud.

Method 2:

Install the camera horizontally and set a threshold at a certain height; points below this threshold are considered as ground.

Debug Visualization Functionality

Visualization information for obstacles can be obtained during debugging by subscribing to the topic /rgbd_pc/obstacle.
Visualization information for ground can be obtained during debugging by subscribing to the topic /rgbd_pc/ground.

Communication Protocol

Input

Topic Name	Type	Description
/camera$/depth/points	sensor_msgs/Image	Point cloud data

Output

Topic Name	Type	Description
/rgbd_pc/obstacle	sensor_msgs::PointCloud2	Point cloud for obstacle detection (debugging)
/rgbd_pc/ground	sensor_msgs::PointCloud2	Point cloud for ground detection (debugging)
/rgbd_pc/laserCloud	sensor_msgs::PointCloud2	Point cloud projected onto a 2D plane

Recommend Projects