Multi-object tracking using semantic analysis and Kalman filter

A generic approach for tracking humans and objects under occlusion using semantic analysis is presented. The aim is to exploit knowledge representation schemes, precisely semantic logic where each detected object is represented by a node and the association among the nodes is interpretated as flow paths. Besides, maximum likelihood is computed using our CWHI technique and Bhattacharyya coefficient. These likelihood weights are mapped onto the semantic network to efficiently infer the multiple possibilities of tracking by the manipulation of ldquopropositional logicrdquo at a time window. The logical propositions are built by formularizing facts, semantic rules and constraints associated with tracking. Currently, we are able to handle tracking under normal, occlusion, and split conditions. The experimental results show that the proposed approach enables accurate and reliable tracking by resolving the ambiguities of online data association under occlusions.