Object Detection Using LiDAR

Overview

This project processes LiDAR point cloud data using the Point Cloud Library (PCL) to detect and classify objects. The pipeline employs:

• Custom RANSAC for planar segmentation
• KD-Tree clustering algorithm for object grouping and identification

The method enables efficient object recognition and spatial analysis, enhancing navigation and environment mapping. The project is developed in C++ within a ROS 2 (Jazzy) environment and visualized using RViz.

Results

Detection and classification of objects in a 3D environment using RANSAC segmentation and KD-Tree clustering is available below:

Repository Structure

Object-Detection-Using-LiDAR/
│── data/pcd/
│   │── data_1/                            # PCD files dataset 1 (highway only cars and trucks)
│   │── data_2/                            # PCD files dataset 2 (highway cars, trucks, bicycles, etc)
│── src/lidar_object_detection/
│   │── include/lidar_object_detection/    # Header files
│   │── src/                               # Source files
│── videos/
│   │── output_lidar_video.gif             # Result Gif (with PCD data_1)
│── CMakeLists.txt                         # CMake build configuration
│── package.xml                            # ROS 2 package metadata
│── README.md                              # Project documentation

Dependencies

Ensure you have the following dependencies installed:

• ROS 2 Jazzy (Robot Operating System 2)
• Point Cloud Library (PCL)
• C++ (CMake build system)
• RViz (for visualization)

Installation & Compilation

Clone the repository into your ROS 2 workspace:

cd ~/ros2_ws/src
git clone https://github.com/your-repo/Object-Detection-Using-LiDAR.git
cd ..
colcon build
source install/setup.bash

Running the Node

Run the ROS 2 node to process LiDAR data:

ros2 run lidar_object_detection lidar_detection_node

Acknowledgements

Special thanks to Udacity for providing the dataset to implement this project.