Classification performances of Mine Hunting Sonar : Theory, practical results and operational applications
Résumé
As low target strength objects can be detected by the new generation of mine hunting sonar systems, more and more non mine like objects are passed on to the classification process. In the past, the mine hunting performance was mainly driven by the performance of the mine detector. With this improved detection capability, the mine hunting system performance is now driven by classification. The classification decision process, based on sonar images including shadows and echoes features, corresponds to the discrimination between MILCO (MIne Like Contact) and NON MILCO. This paper describes the operational incidence of a new theoretical approach used for shadow and image based classification performance evaluation. This approach is named DMC for Delta-Mask based Classification as it assumes the knowledge of specific objects images characteristics, represented with a specific statistical grid named the Delta-Mask. This theory leads to the computation of COC (Classification Operational Characteristics) curves, which give the probability of classification (Pc) and the probability of false classification (Pfc) as functions of sonar characteristics and operational conditions. These curves are computed for several types of sonar systems. Results based on real practical sonar images are shown and operational incidence is deduced.
Origine : Fichiers produits par l'(les) auteur(s)
Loading...