Texture-less object recognition using contour fragment-based features with bisected local regions

摘要

The present invention relates to a method for recognizing objects having bisected local regions and using a contour-division-based feature. The method for recognizing objects having the bisected local regions and using a contour-division-based feature according to the present invention comprises the steps of: being inputted with video data and recognizing the shapes of the objects in the video data; detecting outlines of the objects recognized in the video data and extracting features based on each objects contour division; calculating feature descriptors about the features extracted by the objects contour division in the local regions bisected along a normal line; defining a new pair of descriptors among the feature descriptors about the features; producing a pair of descriptors vector corresponding to the pair of descriptors; executing a spectral matching between the video data and model data extracted with the feature descriptors about the contour-division-based objects; and improving the outcome of the spectral matching.