Negative observations for multiple hypothesis tracking of dynamic extended objects

A novel approach to utilizing negative information to improve multiple hypothesis tracking (MHT) of extended objects is presented. Specifically, a missed detection of a tracked object is treated as an observation of object occlusion, the likelihood of which is described by an occlusion-based sensor detection model. This negative observation is used to update hypothesis weights in a method that is supplementary to any existing MHT framework, as it continues to inform the filter in the absence of traditional measurements. Experimental results are presented demonstrating that integrating this additional interpretation of the available information improves tracking performance.