Bird’s Eye View (BEV) LiDAR Visualization

This project processes raw LiDAR data and visualizes it as a Bird's Eye View (BEV) RGB image using basic preprocessing and point cloud filtering.

Features

• Loads .bin LiDAR files
• Removes irrelevant points based on 3D spatial boundaries
• Projects the LiDAR data into a 2D BEV feature map
• Saves and displays the resulting RGB BEV image

Directory Structure

project/
├── data/
│   └── velodyne/
│       └── 0000000000.bin
├── utils.py
├── config.py
├── main.py
└── output.png

Dependencies

Ensure you have the following Python packages installed:

pip install numpy opencv-python

File Descriptions

• main.py: Main script to generate and display the BEV image.
• utils.py: Contains helper functions for loading and processing LiDAR data.
• config.py: Configuration for BEV grid size, discretization, and 3D boundary limits.

Running the Script

Place a LiDAR .bin file in the data/velodyne/ directory. Then run:

python main.py

This will:

1. Load and preprocess the LiDAR data.
2. Generate a BEV RGB image.
3. Save it as output.png.
4. Display it using OpenCV.

Configuration

Adjust config.py for:

• BEV_WIDTH: Width of the BEV map.
• DISCRETIZATION: Grid resolution.
• boundary: 3D spatial limits for filtering points.

Acknowledgements

This project draws inspiration from and acknowledges the following works:

• kitti2coco by packyan in Python 3.6 and another one. they both convert from kitti to coco and inversibly
• Complex-YOLOv3: For parsing and training on the KITTI dataset, as well as prediction on BEV lidar data. A large part of the codes is berrowed from their excellent work!