Propojení dat z LiDAR senzoru s obrazem kamery pro pokročilou prostorovou interpretaci scény

Abstract

This bachelor thesis focuses on the fusion of LiDAR and camera data for advanced spatial scene interpretation in automotive systems. The main objective is to analyze existing approaches to sensor fusion and to design a custom solution for sensor calibration and integration into a unified framework. An extrinsic calibration method is implemented to project LiDAR point cloud data into the camera image plane. The solution includes both manual calibration using a graphical user interface and an automatic calibration approach based on plane detection and semantic segmentation. The system is implemented in the RTMaps environment using computer vision techniques and YOLO-based neural networks. The result is a functional application capable of real-time sensor fusion visualization and distance estimation of detected objects. The proposed solution is validated on real-world data, and its accuracy and stability are evaluated.