Abstract:

A novel fast local feature description method in real-time image matching is proposed, called Affine Invariant Fast Local Feature (AIFLF). A fast local character descriptor is proposed using affine moment invariants based on normalized gradient value of local neighbor area to the centre point. Simulation experiment show that, for comparative recognition rate, the run time of the proposed AIFLF is obvious less than that of classical SIFT arithmetic. Another matching experiment is conducted, in which the datasets are composed of standard evaluation images, which are taken by mobile robot platform. Performance evaluations show that AIFLF generally outperforms SIFT and SURF, especially in image scale, visual angle, and rotation transformation. Therefore, AIFLF is suitable for real-time image matching application.

Key words: computer application, local feature, affine moment invariants, image matching