Line Matching Method Based on Related Points and Geometric Constraints

Combining related points and straight lines can increase the feature description ability of lines. However, mismatching problems exist for both points and lines when traditional methods are used. To improve line matching accuracy, this paper proposes a line matching method based on related points and geometric constraints. First, a statistical histogram of matched point pair distance ratios is used to eliminate mismatched SIFT feature points. After line extraction, related points are chosen and used to construct affine invariants describing line features. Lines are subsequently coarsely matched based on affine invariant similarities. Finally, an affine transformation model is calculated and lines are finely matched with line angle and distance constraints. Experimental results demonstrate the method is suitable for images with affine transformation and can obtain more matched line pairs with high accuracy than traditional methods. The proposed method considers both local and global geometric line characteristics and obtains good line matching results for images with complex or simple textures.