Classification of Feature Points Required for Multi-Frame Based Building Recognition

The present invention relates to a multi-frame based feature point classifying method for building recognition, the method classifying a feature point necessary for building recognition with one or more multi-frames. The multi-frame based feature point classifying method for building recognition comprises: (a) step of extracting a feature point from each of images of the multi-frames; (b) step of matching each of the images based on extracted feature points to obtain a pair of matched feature points; (c) step of obtaining a homography matrix from the pair of feature points, and classifying the feature points by using the obtained homography matrix; and (d) step of repeatedly performing the (c) step by using the remaining feature points after classifying the feature points. The feature point classifying method is provided to find and classify a feature point which is repeatedly shown by finding an associative relation of the feature points with random sample consensus (RANSAC), thereby increasing accuracy of matching the feature points on the fact that the feature point of a background is shown in a limited image only.