우선 다양한 종류의 data를 종합하여 environment를 조성한다는 것을 이해해야 한다.
DATA Input
1. LIDAR(light detection and ranging)
- Data : Distance Measurements in the form of point clouds.
- Usage : Provides highly accurate spatial information, useful for building detailed maps.
2. cameras
- Data : Vsual information in the form of images or video streams.
- Usage : Used in visual SLAM to extract features from the environment.
3. IMU(Inertial Measurement Unit)
- Data : Acceleration and roatational rates.
- Usage : Helps estimates the robot's movement and orientation.
4. Sonar/Ultrasonic Sensors)
- Data : Distance measurements using sound waves.
- Usage : Useful for short-range distance measurements in simpler environments.
5. GPS(Global Positioning System)
- Data : global position coordinates.
- Usage : Provides rough localization, typically in outdoor environments.
Data Processing
1. Preprocessing
- Filtering : removes noise from the sensor data.
- Calibration : adjust he sensor data to account for any biases or erros.
2. Feature Extraction
- Visual Features : Detecting and describing key points in images(e.g., using SIFT, SURF, ORB).
- Geometric Features : Extracting shapes and structures from point clouds(e.g., lines, planes)
3. Data Association
- Matching : Identifying whether features observed at different times correspond to the same physical feature.
- Tracking : Following features over time to estimate the robot's movement
4. State Estimation
- Localization : Estimating the robot's position and orientation
- Mapping: Constructing a map of the environment.
Data Output
1. Map
- Description : A representation of the environment, which can be 2D or 3D
- Format : bould be in the form of occupancy grids, point clouds or geometic shapes.
2. Robot Pose
- Desciption : The position and orientation of the robot within the map.
- Format : Typically represented as a 3d pose(x, y, z, roll. pitch, yaw)
3. Trajectory
- Description : The path that the robot has traveled over time.
- Format : A sequence of robot poses over time.
4. Environment Features
- Desciption : Key landmarks or features detected in the environment.
- Frmat : points or desciptors that can be used for navigation and further processing.
'On Going > Computer Vision' 카테고리의 다른 글
History of Super Resolution AI (0) | 2024.07.17 |
---|---|
위성영상에 대한 이해 (1) | 2024.07.05 |
SLAM (Simultaneous Localization and Mapping) 관련 개념정리 (0) | 2024.06.18 |
Multi-modal learning (1) | 2023.12.07 |
Landmark localization (1) | 2023.12.05 |
댓글