Feature detection based on epipolar constrains

Affine invariant features detection under large viewpoint change is a challenge problem in photography recent years. With wide application of IMU (inertial measure unit), the elements of exterior orientation can be obtained before aerotriangulation, and then epipolar geometry can be obtained. Besides, in many applications, it is easy to get epipolar geometry, but it is hard to build matching images. Hence the method is proposed to detect fully affine invariant features using Harris detector and match to each other quickly and precisely in the epipolar image, we named it Harris-Epipolar. Feature detection based on epipolar constrains takes full advantage of the characteristics of easy access to the elements of exterior orientation in current conditions and reduces the search space sharply during image matching. It is greatly improved the matching efficiency and accuracy.