Use of Literal Information in Multi-Target Data Association. (Reannouncement with New Availability Information).

摘要

It has been shown that literal information can enhance geolocation information in the multi-target tracking and data association problem. This paper continues previous efforts in establishing a systematic approach to the combination of both types of information using membership functions based upon multiple-valued logic. Filters are established for literal and non-numerical attributes, somewhat analogous to the well-known Kalman filter. The major result, however, is an improvement and clarification of a previous theorem establishing asymptotic forms for the posterior possibility distribution of the unknown data association parameter as information granularity decreases and as inference rule structures become more definitive. The multi-target tracking and data association (or as commonly called, 'correlation') problem still remains the center of much activity and interest. In the past, emphasis was placed upon the use of only geolocation data-i.e., information containing reports on (usually) two- or three-dimensional target positions, together with possible velocities, accelerations and related equations of motion parameters.