Machine Perception

1. 课程信息

  • Location:二教2408
  • Time: Tuesday, 14:00-16:30
  • Teacher:Yongcai Wang
  • Office:Wing building of Science and Technology, 105A
  • Email:ycw@ruc.edu.cn

2. 评分标准

  • 平时作业,占总成绩的40%
  • 课程实验,占总成绩30%
  • 课程论文,占总成绩的30%。

3. 内容简介

本课程主要介绍视觉、惯性感知技术和多传感器融合的机器感知理论与方法,课程内容包括:

  • 贝叶斯滤波,
  • Kalman滤波算法,
  • 定位导航技术,
  • SLAM算法,
  • 图优化算法,
  • 惯性姿态解算,
  • 惯性导航方法,
  • 3D视觉,
  • 视觉里程计,
  • 视觉惯性里程计等。

使得学生了解机器感知的理论、技术基础与学术前沿,特别是通过对相关算法的实现,提高数据处理与智能系统的研发能力。

在每次讲授中,课程内容上将有所调整。


4. 教学安排

讲课内容 周次 学时 课件 参考资料
课程简介 1 1 0. Introduction
贝叶斯滤波理论 1 3 1. Bayesian Filter Probabilistic Robotics _Sebastian Thrun et al.
Kalman滤波,扩展Kalman滤波 2 3 2. Kalman Filter Probabilistic Robotics _Sebastian Thrun et al.
EKF Localization and EKF SLAM 3 3 3. EKF Localization, EKF-SLAM Simulataneous localization and mappingwith the extended Kalman filter‘A very quick guide… with Matlab code!’
Graph SLAM,Gaussian-Newton算法
4 3 4. GraphSLAM Course on SLAM
A Tutorial on Graph-Based SLAM
Notes on Least-Squares and SLAM
Robust Optimizationfor Simultaneous Localization and Mapping
扫描匹配,SVD与ICP算法 5 3 5. Scan Matching Iterative Closest Point (ICP) and other registration algorithms
Efficient variants of the ICP algorithm
3D 刚体运动与姿态变换 6 3 6. 3D Rigid State Transition 3D Rotation in the Space
四元数与三维旋转
惯性导航,零速检测,基于EKF的惯性导航 7 3 7. Inertial Navigation Zero-Velocity Detection—An Algorithm Evaluation
PedestrianTracking withShoe-MountedInertial Sensors
Pedestrian Localisation for Indoor Environments
3D视觉1:2D对极几何,8点法等 8 3 8. 3D vision 1 Maths in 3D computer vision
Visual OdometryPart I: The First 30 Years and Fundamentals
3D视觉2: PnP, 光流,直接法 9 3 9. 3D vision 2 Visual OdometryPart II: Matching, Robustness, Optimization, and Applications
闭环检测、词袋模型、3D建图方法 10 3 10. loop closure and mapping loop closure part in SLAM book
视觉惯性融合里程计,松耦合,紧耦合的视觉里程计方法 11 3 11. visual inertial odemetry (VIO) VINS-Mono: A Robust and Versatile MonocularVisual-Inertial State Estimator, IEEE Trans. on Robotics, 2018
Keyframe-Based Visual-Inertial Odometry Using Nonlinear Optimization
A Semantic Visual SLAM for Dynamic Environment
Visual-Inertial Tightly Coupled Fusionand Nonlinear Optimization for UAVs Navigation
An_Introduction_to_Inertial_and_Visual_Sensing

5. 参考书目

Sebastian Thrun,Probabilistic Robotics _Sebastian Thrun et al.

Peter Corke,《Robotics, Vision and Control

高翔等,《视觉SLAM十四讲

Richard Szeliski,《Computer Vision: Algorithms and Applications》,second Edition

Giorgio Grisetti等,A Tutorial on Graph-Based SLAM

Oliver J. Woodman,An introduction to inertial navigation


6. 课程实验

课程共包括五次实验

1. 设计并实现仿真程序,实现基于Bayes 滤波的移动机器人定位;

2. 根据所提供的数据集实现EKF 定位算法,实现对移动机器人的定位

3. 根据所提供的数据集,实现点云Scan Matching方法,并进行测试

4. 实践Eigen几何模块和Sophus的代码,实现空间3D姿态变换和惯性导航

5. 根据Kitti数据集中的图像,实现2D-2D对极几何,PnP, 光流法,直接法的视觉里程计方法

 


7. 课程论文

1. 课程论文选题lecture report topics

2. 课程论文报告打分表Score Metrics