Automated Threshold Selection for Template-Based Sonar Target Detection
Abstract
This technical report describes a constant false alarm rate (CFAR) detection algorithm along with a threshold selection algorithm derived for target detection in sonar imagery. It is designed for multi-frequency sonars and combines the advantages of template-based detection with environmental adaptation. This combination avoids the numerical estimates of a K-distribution based statistical test, while exploiting geometric features and frequency response differences between the local environment and man-made targets. The detection algorithm is based on our analysis of processed data and derived from a general statistical model. The utility of CFAR algorithms is that the selection of a detection threshold can be made independently of most image variations. However, to account for the widest variety we apply a local adaptive threshold selection mechanism and demonstrate its adaptability.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 01, 2017
- Accession Number
- AD1040463
Entities
People
- Bradley C. Marchand
- Frank J. Crosby