Skip to Main content Skip to Navigation
Conference papers

Non urban driver assistance with 2D tilting laser reconstruction

Abstract : Our researches aim at developing a system which will give assistance to the driver for piloting in rough environments such as off-road or post-crisis environment. The purpose of this approach is to indicate driver where dangers are and how to avoid them. To achieve it, we use 3D reconstruction to detect negative (under the ground) and positive (on the ground) obstacles. This article focuses on the 3D reconstruction. Our approach uses sensors fusion of lidar and stereo cameras and works on a simple strategy. To meet real time perspective and to avoid heavy points cloud matching we use visual odometry to position points cloud in the environment. 3D points cloud is created using a 2D lidar tilting with servomotor on 180◦ to obtain a cheap 3D laser scanner. Visual odometry is brought by Libviso2 (Geiger et al., 2011) using egomotion to determine odometry from stereo images pairs.
This approach is conceived to fit our driver assistance task and not for long time mapping. As a result, after passing by points, they are deleted from created map. This way, the system is not limited by disk space or points cloud computation time in order to meet real time 3D reconstruction. Ground truth acquisitions have been led to test the veracity of such an approach in an outdoor environment. Results show that coherent map are obtained but this fusion is not yet suitable for off-road driving. With some improvements on visual odometry, we could obtain a good 3D reconstruction for low speed and high speed driving.
Document type :
Conference papers
Complete list of metadata
Contributor : Cyril Joly <>
Submitted on : Wednesday, April 19, 2017 - 6:33:03 PM
Last modification on : Wednesday, October 14, 2020 - 3:52:37 AM


  • HAL Id : hal-01510775, version 1


Bruno Ricaud, Cyril Joly, Arnaud de la Fortelle. Non urban driver assistance with 2D tilting laser reconstruction. The 9th International Symposium on Mobile Mapping Technology MMT2015, Dec 2015, Sidney, Australia. ⟨hal-01510775⟩



Record views