Feature Matching Based on Corner Affine Invariant
基于角仿射不变的特征匹配

Different image matching of the same scene is a key problem in computer vision, and is frequently used in three-dimensional object reconstruction,object recognition,image alignment,camera self-calibration and so on. Feature point matching is the most common one among a11 kinds of image matching.To solve the problem of 3-dimensional scene reconstruction, and to improve the performance of present feature point matching, a matching scheme which is invariant to perspective deformation induced by changes in viewpoint is required. This paper proposes a novel algorithm of Feature Match Based on Corner Affine Invariant. It selects corners as extracting feature of the image matching, and these corners are characterized by their orientation and angular width. Through calculating affine invariant, the influence of image stretch, skew, rotation, translation and 1ighting conditions is removed, and by using the epipolar geometry as a matching constraint, those outliers are eliminated too. Consequently we realize the feature matching of image pairs with much difference. And the experimentation shows that the algorithm has high matching accuracy and good matching performance.