Building 3D SIFT point clouds for object recoginition [abstract]
Abstract
In this project, we explored using the Scale Invariant Feature Transform to build three dimensional point clouds for object recognition. To create the point clouds, an object is placed on a lazy susan and pictures are taken at 15 degree increments using a single web camera. The Scale Invariant Feature Transform is then used to match keypoints between images, and a position in three dimensional space is calculated. The SIFT keypoints comprising the point cloud model of the object can then be added to a database for matching.