Object Recognition from Local Scale-Invariant Features
An object recognition system has been developed that uses a new class of local image features. The features are invariant to image scaling, translation, and rotation, and partially invariant to illumination changes and affine or 3D projection. These features share similar properties with neurons in inferior temporal cortex that are used for object recognition in primate vision. Features are efficiently detected through a staged filtering approach that identifies stable points in scale space. Image keys are created that allow for local geometric deformations by representing blurred image gradients in multiple orientation planes and at multiple scales. The keys are used as input to a nearest-neighbour indexing method that identifies candidate object matches. Final verification of each match is achieved by finding a low-residual least-squares solution for the unknown model parameters. Experimental results show that robust object recognition can be achieved in cluttered partially-occluded images with a computation time of under 2 seconds.
Stephen Kiazyk completed his bachelor’s in Computer Science at the University of Manitoba, followed by a master’s degree at the University of Waterloo with a specialization in Computational Geometry. He has since left behind his life as a globetrotting academic to return to Winnipeg, where he writes visual effects and simulation software for the film industry. When he’s not at the computer, he spends his free time practicing Yoga and collecting old video games.