LiDAR Obstacle Detection

Introduction

This project shows how to process raw point cloud data obtained from a LiDAR sensor to perform obstacle detection. In this project, the point cloud processing is done using C++ and Point Cloud Library (PCL). Point cloud segmentation is done using Random Sampling Consensus (RANSAC) algorithm and Euclidean Clustering is used to find clusters in the point cloud data. Finally, a bounding box is created around the detected clusters which show the obstacles detected in the scene. The LiDAR data is pre-processed before applying all the steps mentioned above. Pre-processing is done to downsample the LiDAR data so that obstacle detection can be done efficiently without unnecessarily processing a large number of data points. Voxel grid is used for downsampling the LiDAR data points. Voxel grid allows only one point per voxel which decreases the resolution of the point cloud allowing faster processing. The result of LiDAR point cloud processing is shown in the GIF above.

Project Build Instructions

Ubuntu

git clone https://github.com/sumukhpatil/LiDAR-Obstacle-Detection-Project.git
cd LiDAR-Object-Detection
mkdir build && cd build
cmake ..
make
./environment

Build Dependencies

• cmake version 2.8 or above (click here for installation instructions)
• make version 4.1 or above
• g++/gcc version 5.4 and above
• pcl version 1.2 or above

Point Cloud Library (PCL) Installation

Ubuntu

sudo apt update
sudo apt upgrade
sudo apt install libpcl-dev

Windows

http://www.pointclouds.org/downloads/windows.html

MAC

Install via Homebrew

1. install homebrew
2. update homebrew

   brew update

3. add homebrew science tap

   brew tap brewsci/science

4. view pcl install options

   brew options pcl

5. install PCL

   brew install pcl