Clustering and Vehicle Tracking of vehicles in Pittsburgh with Visualization

This project involves clustering and tracking vehicles using point cloud data from multiple sensors taken by the dataset Argoverse 2. The data is processed to transform coordinates, perform clustering, tracking and visualize the results using Open3D. The original files are not stored, but they have columns: timestamp, yaw, pitch, distance, distance_noise, x, y, z, which contain the coordinates of each point extracted by the LIDAR sensor; x_noise, y_noise, z_noise contain the same data with added noise to simulate the imperfections of a real sensor; object_id, color (R), color (G), color (B).

The file name follows this convention: sensor_X_YY.csv provides the output of sensor X at scan YY.

The file pitt_trajectories.csv contains the ground-truth trajectories of the vehicles, while the file pitt_sensor_positions.csv contains the positions of the sensors.

Table of Contents

• Installation
• Usage
• Project Structure
• Modules
• License

Installation

1. Clone the repository:

   git clone https://github.com/alessandrocapialbi/LiDAR_Cluster_Track
   cd LiDAR_Cluster_Track

2. Create a virtual environment and activate it:

   python3 -m venv venv
   source venv/bin/activate

3. Install the required packages:

   pip install -r requirements.txt

Usage

1. Run the main script:

   python main.py

2. Follow the prompts to enter the number of sensors.

Project Structure

LiDAR_Cluster_Track/
├── data_loader.py
├── main.py
├── clustering.py
├── simulation.py
├── transform_coordinates.py
├── tracking.py
├── sensor_selection.py
├── sensors_data/
├── filtered_sensors_data/
├── sensors_positions/
│   └── pitt_sensor_positions.csv
└── trajectories/
    └── pitt_trajectories.csv
└── output/
    └── point_clouds.csv
    └── bounding_boxes.csv
    └── predicted_trajectories.csv

Modules

main.py

The main script that orchestrates the loading, processing, and visualization of point cloud data.

data_loader.py

Contains functions to load data from files.

clustering.py

Performs DBSCAN clustering on the point cloud data and creates bounding boxes.

simulation.py

Handles the visualization of point clouds and trajectories using Open3D.

transform_coordinates.py

Transforms the coordinates of the point cloud data to a global coordinate system, considering translations and rotations.

kalman_filter.py

Implements a LKF (Linear Kalman Filter) to predict the trajectories of the vehicles.

tracking.py

Tracks vehicles between scans based on bounding box centroids.

sensor_selection.py

Allows the user to select the sensors to use for the tracking.

License

This project is licensed under the MIT License. See the LICENSE file for more details.