Math models 3D to 2D.ppt

Stockman MSU/CSE Math models 3D to 2D Affine transformations in 3D; Projections 3D to 2D; Derivation of camera matrix form Intuitive geometry first Look at geometry for stereo Look at geometry for structured light Look at using multiple camera THEN look at algebraic models to use for computer solutions Review environment and coordinate systems Imaging ray in space General stereo environment General stereo computation Measuring the human body in a car while driving (ERL,LLC) General environment: structured light projection Advantages/disadvantages + can add features to bland surface (such as a turbine blade or face) + can make the correspondence problem much easier (by counting or coloring stripes, etc.) - active sensing might disturb object (laser or bright light on face, etc.) - more power required Industrial machine vision case Industrial machine vision case Develop algebraic model for computer’s computations Need models for rotation, translation, scaling, projection Algebraic model of translation Algebraic model for scaling Algebraic model for rotation Algebraic model of rotation Arbitrary rotation General rigid transformation Moving a 3d object on ANY path in space with ANY rotations results in (a) a single translation and (b) a rotation about a single axis. (Just think about what happens to the basis vectors.) Example: transform points from W to C coordinates Example change of coordinates How to do rigid alignment: perhaps model to sensed data Problem: given three matching points with coordinates in two coordinate systems, compute the rigid transformation that maps all points. Application: model-based inspection Algorithm for rigid 3D alignment using 3 correspondences Algorithm for 3d alignment from 3 corresponding points We assume that the triangles are congruent within measurement error, so they actually will align. Approach: transform both spaces until they align (I) 3d alignment (II) Final solution: rigid alignment Deriving the camera matrix We now combine