Unit
- World coordinate system: A fixed reference frame (often centered at the origin of a scene) that defines the position of all objects in 3D space. It uses three orthogonal axes (X, Y, Z) typically following the right-hand rule, where the Z axis points upward or forward depending on convention.
- Local coordinate systems: Each tracked object has its own frame that moves and rotates relative to the world frame. These are essential for describing the object's orientation independent of its global position.
- Homogeneous coordinates: Representing 3D points as 4-element vectors (x, y, z, w) enables affine transformations (translation, rotation, scaling) through matrix multiplication, simplifying chained operations.
- Transformation hierarchy: Complex motion is built by composing successive local transformations. For example, a robotic arm's end-effector position is derived from the base, through each joint, using concatenated transformation matrices.
- Euler angles vs. quaternions: Euler angles (roll, pitch, yaw) are intuitive but suffer from gimbal lock. Quaternions (a, b, c, d) provide smooth interpolation and no singularities, making them ideal for camera tracking.
- Rotation matrices: 3x3 orthogonal matrices with determinant +1 represent any rotation. They are used in calibration and sensor fusion but are less compact than quaternions.
- Translation vectors: Simple 3D offsets that move an object from one position to another. Combined with rotation, they form a complete rigid-body transformation.
- Scaling and shear: Uniform scaling changes object size, while non-uniform scaling distorts proportions. Shear transforms skew shapes; these are less common in tracking but appear in calibration distortion models.
- Coordinate frame conventions: Different domains use different axis orientations (e.g., Y-up in graphics, Z-up in engineering). Tracking systems must convert between conventions using pre‑defined rotation matrices.
- Quaternion interpolation (SLERP): Spherical linear interpolation between two quaternions produces smooth rotational motion without angular velocity discontinuities, critical for animating tracked camera paths.
- Rigid body transformation: A combination of rotation and translation that preserves distances and angles between points on an object. In tracking, it models the object's pose (position and orientation) in 3D space. Represented as a 4x4 homogeneous matrix, it allows seamless composition of multiple transformations along a kinematic chain.
- Camera extrinsics and intrinsics: Extrinsics define the camera's pose (rotation and translation) relative to the world frame. Intrinsics describe internal parameters like focal length, principal point, and lens distortion. Together, they project 3D world points onto 2D image coordinates, essential for optical tracking systems.
- Degrees of freedom (DoF): A rigid body in 3D space has six degrees of freedom: three translational (X, Y, Z) and three rotational (roll, pitch, yaw). Tracking systems estimate these six parameters to fully describe an object's motion. Some systems (e.g., IMUs) may provide only partial DoF, requiring fusion with other sensors.
Design Principle
Principle: Right-Hand Coordinate System Consistency
Definition: Maintaining a uniform orientation of coordinate axes across all stages of a tracking pipeline to prevent sign errors and misinterpretation of rotations. The cross product of X and Y must always yield Z as per the right-hand rule.
Application: When integrating data from multiple sensors (optical, inertial), convert each sensor's native coordinate frame to the common world frame using a pre‑calibrated transformation. Validate by checking that a known rotation (e.g., 90 degrees around Z) produces consistent results across all sensors. Use a dedicated utility function for frame conversion to avoid coding errors.
Color Palette
Harmony: Triadic for axis visualization
#FF3333X Axis Red
#33FF33Y Axis Green
#3333FFZ Axis Blue
Quick Revision
- World and local coordinate systems define spatial relationships; homogeneous coordinates unify transformations.
- Quaternions avoid gimbal lock and enable smooth interpolation; Euler angles are intuitive but limited.
- Transformation hierarchies compose local motions to model complex structures.
- Consistent axis conventions prevent errors; color-coding axes aids debugging.