3D rekonstrukce okolí vozidla

Abstract

This work describes a system for 3D Reconstruction of the Vehicle Neighborhood. Each part of 3D reconstruction, such as detection and tracking features or triangulation is designed and implemented, and are described their issues. For the detection of the features are used FAST and GFTT detectors. Tracking points between the images is carried out by calculating the optical flow using pyramidal Lucas-Kanade algorithm. For triangulation is used the linear least-square method. Final visualization has a form depth map and interactive visualizer.