Introduction to VSLAM Presentation
출처 : SpaceAnubis, 2013.07
## 1. What is VSLAM
## 2. SfM vs. VSLAM
## 3. Classic Approach : MonoSLAM (EKF-SLAM)
## 4. Current State-of-the-art (2013) : FastSLAM
## 5. Challenge
## 6. DTAM : Dense Tracking and Mapping
1. What is VSLAM
Sensor
- 모노카메라
- 스트레오 카메라
- 키넥트
Feature (a.k.a Landmarks
- point feature
- Geometric feature (eg. lines, planes)
- Featureless
- Object
ROS VSLAM
2. SfM vs. VSLAM
3. Classic Approach : MonoSLAM (EKF-SLAM)
3.1 EKF-SLAM:Mono SLAM (Andrew Davision ICCV 2003)
- Maintains full camera and feature covariance
- Limited to Gaussian uncertainty
- Handheld camera
- Initialized with known target
- Extended kalman filter
Constant velocity
motion model- Image patch features with Active Search
- Automatic Map Measurement
- Particle Filter for initialization of new feature
상세 알고리즘 설명 포함되어 있음, 동영상 참고
3.2 SfM Approach (Nister ICCV 2003)
- RANSAC
- Frame-to-Frame motion only