A robust solution for object recognition by mean field annealing techniques

Object recognition in multi-context scene is one of the very difficult problems to find a robust solution in many applications. The annealed Hopfield networks have been developed to find global solutions of a non-linear system. In the study, it has been proven that the system temperature of MFA is equivalent to the gain of sigmoid function of Hopfield network. In our early work, we developed the hybrid Hopfield network (HHN) on the purpose of fast and reliable matching in the object recognition process. However, HHN does not guarantee global solutions and yields false matching under heavily occluded conditions because HHN is depending on initial states by its nature. In this paper, we present the annealed Hopfield network (AHN) to find a robust solution for occluded object matching problems in multi-context scenery. In AHN, the mean field theory is applied to the hybrid Hopfield network in order to improve computational complexity of the annealed Hopfield network and provide reliable matching under heavily occluded conditions. AHN is slower than HHN. However, AHN provides near global solutions without initial restrictions and provides less false matching than HHN. The robustness of the algorithm is proved by identifying occluded target objects with large tolerance of their features. Also, we present a optimal boundary smoothing algorithm to extract reliable features from the boundary representation of the object heavily contaminated by noise.