Calibrated SAS: Relevance to ATR and Effects of Variability

The proposed effort includes an investigation of calibrated synthetic aperture sonar (SAS) data for use in automated target recognition (ATR) algorithms. Specifically, the research will focus on the role calibrated imagery would play in ATR performance. Currently, the imagery used for ATR is normalized and only yields relative contrast information. The initial work will compare performance against different calibration approaches. Three approaches will likely be compared in the evaluation: 1) a current HF SAS beamforming approach that produces imagery meant to be analyzed by a human operator, 2) a beamforming approach that sets a common mean background level referenced to an arbitrary scattering strength, and 3) a sophisticated calibrated beamforming approach that is capable of estimating absolute target strength. To evaluate performance, calibrated HF SAS ATR algorithms will be coded into the Modular Algorithm Testbed Suite (MATS) and also developed for the advanced synthetic aperture sonar imaging engine (ASASIN). A second thread of the study is to investigate sources of systematic and random uncertainty incalibrated measurements, and understand the degree to which calibrated measurements can be compared across measurement systems, sensing modality (real vs. synthetic aperture), and frequency.