A Neuro-Immune Inspired Computational Framework and Its Applications to A Machine Visual Tracking System

This thesis proposes a bio-inspired computational framework, the Neuro-Immune Network reGulating (NING) framework. In higher vertebrates, the immune system and the nervous system, corporately, maintain a level of internal stable state of the hosts. This internal stable states is known as homeostasis. Similarly, an artificial homeostasis is targeted in this NING framework. There are two networks, an artificial neural network and a novel immuneinspired network. The two networks are organically integrated together bymeans of mutual regulations. Particularly, the immune-inspired network has its distinguishable structures and functions, which differs from any other known artificial neural or immune networks. The NING framework is a framework rather than an algorithm. It can be embodied in engineering applications by applying specific algorithms. In this thesis, a machine visual tracking system is created to characterise and parameterise the NING framework. This demonstration system is able to visually track a morphing target regardless of a non-benign background. This adaptive and robust tracking process is considered as maintaining an artificial homeostasis.