Structure and Motion for Intelligent Vehicles Using an Uncalibrated Two-Camera System

Xinfang Zhang,Jian Chen,Jianqiang Deng,Rong Wang,Wenyi Xiong,Qi Wang
DOI: https://doi.org/10.1109/tie.2022.3163509
IF: 7.7
2022-01-01
IEEE Transactions on Industrial Electronics
Abstract:In this article, the structure and motion of a moving object are asymptotically estimated by an uncalibrated vehicle-mounted two-camera system under uneven road condition. Concurrent learning techniques are utilized to compensate for the cameras’ extrinsic parameters, and a time-varying gain is introduced to accelerate the estimation. Meanwhile, geometric relationships are used along with the visual information to obtain depth and 3-D coordinates of a feature point located on the moving object. Homography techniques are utilized to construct a measurable auxiliary state whose dynamics is related to the moving point’s velocity, which is further obtained by a nonlinear observer. Lyapunov-based analysis is conducted to prove the asymptotic convergence of the cameras’ extrinsic parameters as well as the moving point’s depth and velocity. The proposed method is applicable for different sensor configurations and road conditions as long as proper linear parameterization is done for the vision-based dynamics. Both simulations and experiments are conducted to verify the effectiveness and performance of the proposed method.
What problem does this paper attempt to address?