Creating device and how to create a three-dimensional object recognition for image database

摘要

While reducing the memory capacity by performing sift the local feature quantity registered in the three-dimensional object recognition image database, to suppress a decrease in the recognition rate. A plurality of images seen from different viewpoints three-dimensional object is input, the steps, each of which represents a characteristic vector extracting local features of each image, and a group of consisting of feature vectors, each set is next a step of generating a plurality of partial spaces for generating a set representing local features viewed same portion of the object from the perspective of the set of fit, representing the characteristics of each set, and adds the ID of the object in each subspace When a step of registering in the database the image viewed from a certain viewpoint and an object is given as a search query, extracts the feature vectors that represent local features of the search query, to be most similar to the feature vector determining each subspace, performs aggregation processing for the object ID added to each subspace, creating a three-dimensional object recognition image database comprising the steps of obtaining an object that is most similar to the search query.