Homography Transformation
A 3×3 matrix that describes the projection relationship of the same plane across two viewpoints. It is commonly used for document rectification, bird's-eye view transformation and planar augmented reality, and can be estimated from at least four pairs of matched points.
Timeline (last 90 days)
Recovering top-down coordinates from an obliquely shot single-camera frame is essentially a homography problem, requiring finding a 3×3 projection matrix to map pixel coordinates to the ring plane's real coordinate system.
Homography belongs to projective transformation, which is a superset of affine transformation; affine transformations preserve parallelism between lines, while projective transformations allow parallel lines to meet at a point (vanishing point), accurately describing the perspective geometry of camera imaging.
The homography matrix H has 9 elements, but due to the scale invariance of homogeneous coordinates, it actually has only 8 degrees of freedom; therefore, at least 4 pairs of non-collinear corresponding points are needed to solve it uniquely.
Homography has a fundamental prerequisite constraint: the target must lie on the same plane; when the target deviates from that plane, the mapping result will exhibit significant error, and a complete 3D reconstruction (involving depth estimation or multi-view geometry) is required to obtain accurate positions.
Homography is a mathematical tool describing the projective relationship between two planes, using a 3×3 matrix H to establish a one-to-one mapping between image pixel coordinates (x, y) and ground truth coordinates (X, Y).
Calibration reference points should be as widely spread out as possible and cover the entire region of interest, avoiding four points that are too clustered or nearly collinear; using more than 4 points for least-squares fitting improves robustness; it is recommended to use a laser rangefinder or RTK-GPS to accurately measure reference point coordinates.
Homography transformation does not require a complete camera intrinsic calibration, as it directly establishes a mapping relationship between two planes, making it more lightweight and easier to deploy than a full 3D reconstruction approach.
In pedestrian ground localization, the homography transformation should use the bottom center point of the detection bounding box (foot position) rather than the head, because only the feet actually lie on the ground plane.
All Facts (8)
Recovering top-down coordinates from an obliquely shot single-camera frame is essentially a homography problem, requiring finding a 3×3 projection matrix to map pixel coordinates to the ring plane's real coordinate system.
50%UnverifiedHomography belongs to projective transformation, which is a superset of affine transformation; affine transformations preserve parallelism between lines, while projective transformations allow parallel lines to meet at a point (vanishing point), accurately describing the perspective geometry of camera imaging.
50%UnverifiedThe homography matrix H has 9 elements, but due to the scale invariance of homogeneous coordinates, it actually has only 8 degrees of freedom; therefore, at least 4 pairs of non-collinear corresponding points are needed to solve it uniquely.
50%UnverifiedHomography has a fundamental prerequisite constraint: the target must lie on the same plane; when the target deviates from that plane, the mapping result will exhibit significant error, and a complete 3D reconstruction (involving depth estimation or multi-view geometry) is required to obtain accurate positions.
50%UnverifiedHomography is a mathematical tool describing the projective relationship between two planes, using a 3×3 matrix H to establish a one-to-one mapping between image pixel coordinates (x, y) and ground truth coordinates (X, Y).
50%UnverifiedCalibration reference points should be as widely spread out as possible and cover the entire region of interest, avoiding four points that are too clustered or nearly collinear; using more than 4 points for least-squares fitting improves robustness; it is recommended to use a laser rangefinder or RTK-GPS to accurately measure reference point coordinates.
50%UnverifiedHomography transformation does not require a complete camera intrinsic calibration, as it directly establishes a mapping relationship between two planes, making it more lightweight and easier to deploy than a full 3D reconstruction approach.
50%UnverifiedIn pedestrian ground localization, the homography transformation should use the bottom center point of the detection bounding box (foot position) rather than the head, because only the feet actually lie on the ground plane.
50%