open3DCV a lightweight Multiple View Geometry library

• Home
• Installation
• Dependencies
• Building
• Documentation
• Keypoint
• Math
• Visualization
• Triangulation
• Transform
• Robust Estimator
• Pose Estimator
• Matching
• IO
• Image
• Camera Models
• Tutorial
• Open3DCV Datatypes
• Bundle Adjustment using Ceres
• Structure from Motion
• Bibliography
• Bibliography
• Developers
• License
• Contributing
• Changelog

Fundamental Matrix

class Fundamental_Estimator

This class is the sub-class of class Param_Estimator, which is a base class for various estimators that can be used by the Ransac class. Please visit the RANSAC page for details.

Seven-point algorithm

void fund_seven_pts (const std::vector<Vec2f>& x1, const std::vector<Vec2f>& x2, vector<Mat3f>& F);

method that estimates fundamental matrix using seven corresponding pairs, which may return 1 or 3 results.

Eigen-point algorithm

void fund_eight_pts(const std::vector<Vec2f>& x1, const std::vector<Vec2f>& x2, Mat3f& F);

method that estimates fundamental matrix using eight corresponding pairs. The fundamental matrix is estimated up to a scale.

Robust estimation using RANSAC

void estimate(std::vector<DMatch>& data, std::vector<float>& params);

method inherited from class Param_Estimator, which is used within the Ransac class to estimate parameters using either seven-point or eight-point algorithm. I suggest to use seven point algorithm since fewer data requires fewer iterations.

void ls_estimate(std::vector<DMatch>& data, std::vector<float>& params);

method that uses eight-point algorithm to estimate the least squares solution. This method is invoked at the end of RANSAC once the inliers have been detected.

int check_inlier(std::pair<DMatch>& data, std::vector<float>& params);

method used to see if the data fits the estimated model.

Rigid pose from Essential matrix

Rt_from_E.h

This file implements the algorithm to estimate relative pose (rotation and translation) of two views from Essential matrix. This method is used extensively in two-view SfM to estimate relative pose of two viewpoints.

void Rt_from_E(const Mat3f& E, std::vector<Mat3f>& R, std::vector<Vec3f>& t);

method that returns two rotation matrices and two translation vectors due to ambiguity.

void Rt_from_E(Pair& pair);

class Pair is a data structure that stores various information of two views, including relative pose, matching points, and so on. This method invokes the void Rt_from_E(const Mat3f& E, std::vector<Mat3f>& R, std::vector<Vec3f>& t) method, and then eliminates ambiguity by ensuring that the reconstructed structure is in front of the two views.

Essential Matrix

class Essential_Estimator

Homography

class Homography_Estimator