Cones Perception via LiDAR Sensor

A ROS package for detecting and classifying traffic cones in LiDAR point cloud data using PCL (Point Cloud Library) and clustering algorithms.

🚀 Features

• Real-time cone detection from LiDAR point clouds
• Color classification of detected cones (yellow, blue, orange, unknown)
• Ground removal preprocessing for improved accuracy
• Multiple LiDAR support (custom, FSAI, simulation)
• Configurable parameters for different environments
• Euclidean clustering for robust cone segmentation
• Voxel grid filtering for performance optimization

📋 Requirements

• ROS (Robot Operating System) (tested on Noetic)
• PCL (Point Cloud Library)
• sensor_msgs
• tf
• std_msgs

🛠️ Installation

1. Clone this repository into your catkin workspace:

cd ~/catkin_ws/src
git clone https://github.com/dmn-sjk/cones_perception.git

2. Build the package:

cd ~/catkin_ws
catkin_make

3. Source your workspace:

source ~/catkin_ws/devel/setup.bash

🚀 Quick Start

Basic Usage

roslaunch cones_perception cones_perception.launch

Advanced Usage

# With ground removal enabled
roslaunch cones_perception cones_perception.launch ground_removal:=true

# With color classification
roslaunch cones_perception cones_perception.launch classify_colors:=true

# For different LiDAR types
roslaunch cones_perception cones_perception.launch used_lidar:=fsai
roslaunch cones_perception cones_perception.launch used_lidar:=simulation

📊 Published Topics

The package publishes detected cones to the following topics:

• /cones_perception/cones_cloud_unknowns - Unclassified cones
• /cones_perception/cones_cloud_yellows - Yellow cones
• /cones_perception/cones_cloud_blues - Blue cones
• /cones_perception/cones_cloud_oranges - Orange cones

⚙️ Configuration

Detection Parameters

Key parameters can be configured in config/cones_detection_params_*.yaml:

• distance_treshold_min/max: Detection range (0.7-7.0m)
• min_cluster_size/max_cluster_size: Cluster size limits (3-50 points)
• voxel_filter_leaf_size_*: Voxel grid resolution (0.04m)
• level_threshold: Ground level filtering (-0.5m)
• angle_threshold: Detection angle range (90°)

🎯 Algorithm Overview

1. Point Cloud Preprocessing: Voxel grid filtering and ground removal
2. Spatial Filtering: Distance and angle-based point filtering
3. Clustering: Euclidean clustering to group cone points
4. Centroid Calculation: Computing cone positions and orientations
5. Color Classification: Optional color-based cone categorization
6. Publishing: Outputting classified cone point clouds