PLiNC algorithm: pattern location in noisy contexts

The problem that is addressed can be likened to that of detecting the position of a random dot leopard continually and incoherently changing its spots against a continually and also incoherently changing similar (but not stochastically identical) spotted background. An algorithm has been developed, which can locate near-rigid targets consisting of spatially separated patches presented against a background of very similar texture. The patches and background are drawn from different distributions, and renewed at every time step. The algorithm works by efficiently integrating weak measurement information time, converging rapidly to a hypothesis associated probability. The measurements and associated `uncertainties´ are obtained using the Forstner corner and circular feature algorithm. The results of the algorithm appear superior to the human ability to detect these sort of targets in noise textures. The mathematics elaborating some formal constraints on the performance of the algorithm is presented