Please use this identifier to cite or link to this item:
Title: Stereo vision for obstacle detection in autonomous vehicle navigation
Keywords: Autonomous Navigation, Stereo vision, Dense Disparity, Ground Plane Modeling, Obstacle Detection, Semi-structured Terrains
Issue Date: 16-Aug-2010
Citation: SAMEERA KODAGODA (2010-08-16). Stereo vision for obstacle detection in autonomous vehicle navigation. ScholarBank@NUS Repository.
Abstract: Autonomous navigation has attracted an unprecedented level of attention within the intelligent vehicles community over the recent years. In this work, we propose a novel approach to a vital sub-problem within this domain, obstacle detection. In particular, we are interested in outdoor rural environments consisting of semi-structured roads and diverse obstacles. Our autonomous vehicle perceives its surroundings with a passive vision system: an off-the-shelf, narrow baseline, stereo camera. An on-board computer processes and transforms captured image pairs to a 3D map, indicating the locations and dimensions of positive obstacles residing within 3m to 25m from the vehicle. The accuracy of stereo correspondence has a direct impact on the ultimate performance of obstacle detection and 3D reconstruction. Therefore, we carefully optimize the stereo matching algorithm to ensure that the produced disparity maps are of expected quality. As a part of this process, we supplement the stereo algorithm by implementing effective procedures to get rid of ambiguities and improve the precision of output disparity. The detection of uncertainties helps the system to be robust against adverse visibility conditions (e.g., dust clouds, water puddles and over exposure), while sub-pixel precision disparity enables more accurate ranging at far distances. The first and the most important step of the obstacle detection algorithm is to construct a parametric model of the ground plane disparity. A large majority of methods in this category encounter modeling digressions under direct or indirect influence of the non-flat ground geometry, which is intrinsic to semi-structured terrains. For instance, the planar ground approximation suffers from non-uniform slopes and the v-disparity algorithm is prone to error under vehicle rolling and yawing. The suggested ground plane model on the other hand is designed by taking all such factors into consideration. It is composed of two parameter sets, one each for the lateral and longitudinal directions. The lateral ground profile represents the local geometric structure parallel to the image plane, while the longitudinal parameters capture variations occurring at a global scale, along the depth axis. Subsequently an obstacle map is produced with a single binary comparison between the dense disparity map and the ground plane model. We realize that it is unnecessary to follow any sophisticated procedures, since both inputs to the obstacle detection module are estimated with high reliability. A comprehensive evaluation of the proposed algorithm is carried out using data simulations as well as field experiments. For a large part, the stereo algorithm performance is quantified with a simulated dense disparity map and a matching pair of random dot images. This analysis reveals that our stereo algorithm is only second to iterative global optimization amongst the compared methods. A similar analysis ascertains best suited procedures and parameters for ground plane modeling. The ultimate obstacle detection performance is assessed using field data accumulated over approximately 35km of navigation. These efforts demonstrate that the proposed method consistently outperforms both planar ground and v-disparity methods.
Appears in Collections:Master's Theses (Open)

Show full item record
Files in This Item:
File Description SizeFormatAccess SettingsVersion 
S.KODAGODA_HT070273N_M.Eng Thesis.pdf7.24 MBAdobe PDF



Page view(s)

checked on Apr 18, 2019


checked on Apr 18, 2019

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.