Neuro-fuzzy techniques for image tracking

Abstract

This work presents the application of neuro-fuzzy techniques to develop a rule-based fuzzy system as an efficient and robust approach for object tracking based on sequences of video images. The fuzzy system is a new approach for data association problems in video image sequences, which uses JPDA formulation adapted to cope with video data peculiarities. The neurofuzzy learning algorithm uses examples extracted from a surveillance system of airport surface, including situations with very closely spaced objects (aircraft and surface vehicles moving on apron). The learned fuzzy system ponders update decisions both for the trajectories and shapes estimated for targets with the set of image regions (blobs) extracted from each frame. The inputs of the neuro-fuzzy system are several numeric heuristics, describing the quality of gated groups of blobs and predicted tracks, and outputs are confidence levels used in the update process. Rules are aimed to generate the most appropriate decisions under different conditions, emulating the reasoned decisions taken by an expert, and have been learned by the neuro-fuzzy system. The system performance with real image sequences of representative ground operations is shown at the end. © Springer-Verlag Berlin Heidelberg 2003.