Abstract:Pose estimation is a crucial technique to achieve robot localization using natural features for an RGB-D vision navigation system. However, it still suffers from the impoverishment of robustness and accuracy in complex workspaces. A robust real-time pose estimation approach using hybrid point and twin line reprojection is proposed to allow accurate localization for a vision- based robot under various challenging scenarios, including low texture, low structure, occlusion, and viewpoint and illumination changes. First, an improved adaptive threshold (IAT)-based oriented fast and rotated brief (ORB) feature detection method is developed to extract sufficient IAT-ORB point features that are matched by a multistage K-nearest neighbor (KNN)-ratio-test criterion (RTC)-cosine similarity (CS)-random sample consensus (RANSAC) (KRCR) refinement feature matching strategy to generate reliable point matches. Then, a robust EDLines line feature matching method is developed by integrating dynamic matching (DM) and median absolute deviation (MAD) to obtain fine line matches. Furthermore, the hybrid point reprojection errors are built from the fusion of 3-D-2-D and 3-D-3-D point pairs, and the twin line reprojection errors are constructed by integrating virtual right-eye lines generated by depth measurement, which both are incorporated into a novel unified errors' optimization model to achieve optimal motion estimation by minimizing the reprojection errors of points and lines concurrently. Finally, the extensive experiments verify the robustness, accuracy, and real-time performance of the proposed approach on the public datasets and our real indoor scenarios. The quantitative results demonstrate that our proposed approach gains more accurate and stable localization than other state-ofthe-art methods under challenging environments.

Scene Modeling and Autonomous Navigation for Robots Based on Kinect System

A Computer Vision for Navigation of Mobile Robots

Active Planning of Robot Navigation for 3D Scene Exploration

A Framework for Multi-Session RGBD SLAM in Low Dynamic Workspace Environment.

Simultaneous Intrinsic And Extrinsic Calibration Of A Visual-Odometric Sensor System

A Depth Estimation Framework Based on Unsupervised Learning and Cross-Modal Translation

Visual Odometry Based 3D-Reconstruction

Indoor Objects 3D Modeling Based on RGB-D Camera for Robot Vision

Robust Depth-Aided RGBD-Inertial Odometry for Indoor Localization

RGB-D Vision Navigation Via Motion Saliency Measurement and Twin Reprojection Optimization in Complex Dynamic Scenes

Real-time Multi-view Omnidirectional Depth Estimation System for Robots and Autonomous Driving on Real Scenes

Enhancing Visual Odometry with Estimated Scene Depth: Leveraging RGB-D Data with Deep Learning

Indoor Environment RGB-DT Mapping for Security Mobile Robots.

Visual Odometry via Hybrid Motion Estimation Method

Multi-robot Collaborative SLAM and Scene Reconstruction Based on RGB-D Camera

Visual Odometry and 3D Mapping in Indoor Environments

Robust Pose Estimation via Hybrid Point and Twin Line Reprojection for RGB-D Vision Navigation

RGB-D Inertial Odometry for Indoor Robot via Keyframe-based Nonlinear Optimization

A Two-Stage Monocular Vision Detection Method for 6D Pose Estimation in Multi-Heterogeneous Robot Systems

A Hybrid Approach to Real-Time Robotic Visual Navigation: Integrating Detection and Scene Segmentation

A Human–Robot Collaborative System for Robust Three-Dimensional Mapping