Object Recognition Tools for Educational Robots

SIFT (Scale Invariant Feature Transform) [1] features, developed by David G. Lowe, have been found to be highly robust to translations, rotations and scaling, and have been the solution of choice for many when dealing with problems of robotic vision and object recognition. Object recognition using SIFT features involves detection of SIFT features in an image, and matching those features against features in a structured database of object images. However, to the best of our knowledge, there are currently no open-source tools to conveniently perform these tasks. This research aims to develop SIFT-based object recognition tools for use by students in undergraduate robotics courses. The tools will allow students to gain a basic understanding of SIFT, but also abstract away from the actual implementation. Hence, students will be able to focus on solving higher level robotics problems.