Algorithms for the Fusion of Two Sets of (Sonar) Data
Abstract
In this report we study different methods to combine sonar contacts as observed by two sonars. The sonar contacts are given realistic position errors, which makes their association non-trivial. First, for a single pair of contacts the most-likely position of the true underlying target position is derived. Based on this, the probability that two observed contacts originate from a single object is calculated. Based on these theoretical derivations, different association methods are evaluated using simulations, in which both targets-of-interest and false alarms are inserted. It is concluded that an `OR'-fusion of the two sets of sonar contacts gives a much better performance than an `AND'-fusion; the latter induces severe losses. The results are insensitive to the number of targets inserted, to the exact magnitude of the position errors, and to the amplitude distribution of the targets-of-interest.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 2006
- Accession Number
- ADA470758
Entities
People
- C. A. Van Moll
- P. A. De Theije