Trifocal Tensor's Application And Reallizition To Get Features In Image Sequences

The feature tracking in images is the preconditions of 3D reconstruction, motion tracking and image segmentation, it is important in digital photogrammetry and computer vision. This paper mainly researches the application that one uses the trifocal tensor to get feature points and feature lines.The trifocal tensor is the mathematical expression of trifocal geometry, it has strict constraints to correspondence points and lines in three images. It can restrict the object correspondence points to one point and the object correspondence lines to one line. The trifocal tensor can eliminate the wrong matches while the epipolar geometry algorithms of feature point and line match can not find them.This paper researches denoise, tracking of feature points and feature lines and matching of feature points and feature lines in digital images. Thereinto, the problem of matching feature points and feature lines introduce the trifocal tensor as the constraint condition. In the end, the paper designes a system of getting image features. Experiments demonstrate that the system can preferably get feature points and feature lines.