Three-dimensional Reconstruction Of The Match

Different image matching of the same scene is a key problem in computer vision, and is frequently used in 3-D model reconstruction , object recognition, "image alignment, camera self-calibration and so on. To solve the problem of 3-D scene reconstruction, and to improve the performance of present matching technique, this paper enter upon with the feature matching techniques, make detailed analysis on some famous algorithm of feature extraction and feature matching. In the research of edge feature extraction, we propose an improved Canny operator based on the transition region and dynamic threshold, which implements the function of automatic selection of the double threshold and can detect more details than before. In the research of corner feature extraction, we propose an improved SUSAN operator in order to solve the problem of the precise location and changeless threshold. The improved operator is based on the theory that a corner point must be connected with the corner where it is located, and it also uses the method of self-adaptable threshold. In the research of feature matching, we select corner point which has discussed before as extracting feature of the image matching, and then we proposes a novel algorithm of Feature Match Based on Corner Affine Invariant. Through calculating affine invariant, the influence of image stretch, skew, rotation is removed, And the experimentation shows that the algorithm has high matching accuracy and good matching performance.