Introduction

This example shows the implementation of a Intel Realsense D455 camera inside a simple virtual world made in Gazebo. The camera publishes it's pointcloud and image to ROS. One C++ example node detects objects in the pointcloud and publishes these as markers. Another Python example node detects the orange cones in the images and marks these in the image that is published by the node. This image, together with the pointcloud and the marker detections is visualized in Rviz.

Virtual environment in Gazebo

Pointcloud	Image

Requirements

This example is made using ROS Noetic on Ubuntu 20.04 and there is no garuantee that it works with other distributions. Both OpenCV and PCL libraries are needed for the detection nodes.

Install

To install the packages follow:

mkdir catkin_ws
cd catkin_ws
git clone https://github.com/thijs83/Virtual-Intel-Depth-Camera.git
cd Virtual-Intel-Depth-Camera/
catkin_make

Get started

To launch the whole setup:

source devel/setup.bash
roslaunch realsense_d455 test_environment.launch

This will launch Gazebo and Rviz, and it will start the python and c++ node.

Packages

realsense_d455

This package contains all models to setup the virtual environment and publishing the camera output to ROS

pointcloud_filter

This package contains an example pointcloud filter in c++ that detects objects in the pointcloud retrieved from the camera.

image_filter

This package contains an image filter example using a python script. The filter searches for the colors of the cone (white/orange) and then adds a contour around the detections

References

Two main github repositories are used to implement the virtual D455 camera: pal-robotics/realsense_gazebo_plugin IntelRealSense/realsense-ros