Research On The Line Feature Matching Method Under Multiple Constraints

In the image line features contain more information than point features, at the same time,more easily than surface features in extraction and descirption, therefore, in this paper, take linefeature matching as the key point. Put forward the line feature matching method under multipleconstraints and has a reliable matching result，providing a good foundation for3d reconstruction.The main study content of this paper as follows:Edge detection: Through the contrast analysis of the advantages and disadvantages of eachalgoirthm and in the light of the Canny operator shortcomings and the insufficiency, is proposedto improve the Canny operator based on the dynamic double threshold edge detection method ofimage edge extraction;Line extraction; In the phase grouping method, this paper improves the calculation methodof phase gradient，uses weighted least squares fitting lines and gets more accurately lines;In theHough transform method, this paper has a deal with neighborhood processing and adaptivemedian filtering, two value processing，recording edge point coordinates, embedding resetmethod to detect peak, setting threshold to remove a connection lines and pseudo lines，recordingendpoint position of lines, fitting lines with the least squares, and gets the highlighted andsignificant lines of image atfer the improvement.Line matching: This paper puts forward the method of line matching based on multipleconstraints.This method uses multiple constraint conditions such as Homography matrix,Epipolar line，gray similarity to constraint line matching and completes thick matching. Thenuses consistent constraint to eliminate mistakes in matching line, and realizes the essence of linematching. The result shows that this method has an certain applicability and solves the problemof linear fracture and keep out effectively.