Real-time 3D vehicle reconstruction with 1D laser scanner and monocular camera

Abstract

Recently, vehicle classification has become very important in the field of intelligent transportation. The methods based on 3d model can play a better role for it contains more information from any viewing angles than 2d images. This paper presents a method to reconstruct the three-dimension model of vehicles in real-time when they get through our special device which is composed of two parts: a laser scanner and a camera. The main algorithm can be summarized as two parts: data preparation with mixed calibration and 3d model creating with data synthesis. The 3d vehicle model consists of three parts: vertices, texture, and triangle connection structure. The final result is demonstrated that our method can reconstruct the 3d vehicle model fast and accurately. © 2014 Springer-Verlag Berlin Heidelberg.