Robust Tracking of Humans by Intelligent Multi-modal Fusion of Visual Cues using Machine Learning

This document formulates a technical report proposing a methodology to improve the robustness of vision based human tracking systems in real-world scenarios. We propose a system consisting of 2 stages, 1. a vision based human tracking system using multiple visual cues, 2. an intelligent multi-modal fusion module, using machine learning techniques, to determine the right weights for each visual modality for the operating environment of the system. The function of the second stage is to perform on line analysis of parameters in the current scene that influences the performance of the tracker. Depending on this analysis, optimal weights are generated for each visual modality, indicating its contribution in the current scene. With such a fusion module we intend to boost the robustness of the system.     «

This document formulates a technical report proposing a methodology to improve the robustness of vision based human tracking systems in real-world scenarios. We propose a system consisting of 2 stages, 1. a vision based human tracking system using multiple visual cues, 2. an intelligent multi-modal fusion module, using machine learning techniques, to determine the right weights for each visual modality for the operating environment of the system. The function of the second stage is to perform on...     »