An Object Detection and 3D Reconstruction Approach for Real-Time Scene Understanding

Abstract

This paper presents the theoretical and practical aspects of a 3D object reconstruction approach. The reconstruction process involves the usage of a pair of stereo-rectified images acquired from two digital cameras. The imaged object is rebuilt in a virtual 3D space with the help of internal and external camera parameters obtained from the camera calibration algorithm. For the 2D image detection of the object of interest, a color segmentation method is applied, followed by triangulation for estimating the object-camera distance.