Papers by Author: Chun Xia Zhao

Paper TitlePage

Authors: Wu Zhou, Chun Xia Zhao, Mian Hao Zhang
Abstract: When Simultaneous Localization and Map Building is carried out in complex environments, reduction of computational complexity is a key problem. With a view to the high computational complexity of particle filter, a SLAM solution named ‘Fast Kalman SLAM’ is introduced. Adopting the ‘decomposition’ idea in the FastSLAM algorithm, Fast Kalman SLAM factors the joint SLAM state into a path component and a conditional map component. The robot pose is estimated recursively with Mean Extended Kalman Filter (MEKF) or Unscented Kalman Filter (UKF), while the map with Extended Kalman Filter (EKF). Simulative experiments are carried out to evaluate the performance of the presented algorithm. And Simulation analysis is made for the presented algorithm. The experimental results indicate that the new algorithm reduces computational complexity greatly and ensures estimation accuracy at the same time.
Authors: Cai Ling Wang, Chun Xia Zhao, Jing Yu Yang
Abstract: A high accuracy rotation angle estimation algorithm based on Local Upsampling Fourier Transform (LUFT) is developed in this paper. The LUFT uses a hierarchical strategy to estimate the rotation, which consists of a transformation of rotation to translation, a fast coarse rotation estimation and a robust refinement stage as well. The coarse rotation is estimated through the conventional Phase Only Correlation (POC), then, it is refined by the resampling technique within a local neighborhood in frequency domain. Furthermore, as will be shown in many experiments, the LUFT can achieve high accuracy rotation estimation, where the accuracy is tunable to some extent. Specially, it is efficient and robust to noise.
Authors: Cai Ling Wang, Chun Xia Zhao, Jing Yu Yang
Abstract: A robot monocular localization method based on 4-DOF ego-motion model and RANdom SAmple Consensus (RANSAC) in country road environment is introduced in this paper. The algorithm uses as input only those images provided by a single camera mounted on the roof of the vehicle. Given the image motion, the stable localization of vehicle is computed by the 4-DOF ego-motion model and RANSAC algorithm, which can efficiently estimate the instantaneous steering angle and traveling distance even if the vehicle runs on bumpy road. The proposed algorithms are successfully validated on videos from an automotive platform. Furthermore, the localization results from our method are compared to the ground truth measured with a GPS/INS joint system, and the trajectory recovered from the estimated localization parameters is superimposed onto a digital map. These experimental results indicate that the algorithms are efficient and encouraging.
Showing 1 to 3 of 3 Paper Titles