COMPUTER VISION SYSTEM AND METHOD EMPLOYING ILLUMINATION INVARIANT NEURAL NETWORKS

摘要

Utilize normalized crosscorrelation (NCC) to measure object is classified, with two images that relatively under the condition of illumination unevenness, obtained.Input pattern is classified, to distribute experimental tag along sort and numerical value.Input pattern is distributed to the output node that in radial primary function network, has largest classification value.If input pattern and image (the being called node image) both who is associated with this node have uniform illumination, then accept this node image and probability and be set to be higher than user-defined threshold value.If test pattern or node image are inhomogeneous, then do not accept this node image and classification value is remained the numerical value that sorter distributes.If test pattern and node image are inhomogeneous, then use the NCC measurement and classification setting is the NCC value.