Pseudo-LiDARs with Stereo Vision

For autonomous drones, accurately detecting obstacles and measuring distances is crucial. However, traditional LiDAR LiDARs are expensive and heavy, making them impractical for drones. We need a lighter and cheaper alternative to ensure safe and efficient drone deliveries.

This project offers a solution using stereo computer vision techniques for per-pixel depth estimation, replacing LiDAR sensors. By using block matching and Stereo Global Block Matching (SGBM) algorithms, we accurately determine depth information. Our experiments show that this approach is significantly cheaper than LiDAR but still provides reliable depth estimation.

Overall, integrating pseudo-LiDAR systems presents a cost-effective solution for obstacle detection in drone deliveries. This allows drones to avoid obstacles, plan paths, and understand their environment safely and efficiently.