A Hierarchical Sensing Approach for Efficient Motion Tracking and Analysis in Advanced Surveillance

Motion tracking and analysis is an important topic to study in automated human tracking for surveillance and many other applications. This proposal aims to study and develop a systematic and hierarchical motion sensing approach for advanced surveillance and the related applications. We propose a flexible and generic non-isomorphic sensing model for this purpose. The rational of the new sensing paradigm will be investigated by exploring the physically implemented compressive sampling with the visibility modulated pyroelectric infrared (PIR) sensor arrays. We will study the issues of coverage scalability and sensing efficiency of Boolean compressive infrared sensing. In particular, the hierarchical Boolean compressive infrared sensing will be investigated for achieving scalable visibility coverage and fulfilling multilevel sampling resolution requirements. The sensing efficiency analysis and design will be studied for the compressive infrared sensing with structured sparsity, which can benefit the development of low-cost and lightweight, compact and modularized PIR sensor arrays. To track motions in both narrow and wide fields, we will study multilevel motion tracking based on motion sensing at multilevel granularities. This includes the tasks of exploring lightweight methodologies for data-to-object association and motion inference with hierarchical compressive infrared sensing model. Based on the multilevel motion tracking, synergistic motion analysis will be pursued to integrate the global and local motion inferences from complementary perspective, which can contribute to resolving the ambiguity due to grouping interactions.