Towards a Computational Model for Object Recognition in IT Cortex

摘要: There is considerable evidence that object recognition in primates based on the detection of local image features intermediate complexity are largely invariant to imaging transformations. A computer vision system has been developed performs using with similar properties. Invariance translation, scale and rotation achieved by first selecting stable key points space performing feature only at these locations. The measure gradients a manner modeled response complex cells primary visual cortex, thereby obtain partial invariance illumination, affine change, other distortions. used as input nearest-neighbor indexing method Hough transform identify candidate matches. Final verification each match finding best-fit solution for unknown model parameters integrating consistent parameter values. This procedure provides serial process attention human integrates belonging single object. Experimental results show this approach can achieve rapid robust cluttered partially-occluded images.