Calibrated SAS: Relevance to ATR and Effects of Variability
Abstract
Current high-frequency synthetic aperture sonar (SAS) imaging systems and processing software have been largely successful in producing high-quality imagery in a variety of oceanographic and seafloor environments. In most high frequency real aperture and SAS systems, the image pixel values have been adjusted for optimal viewing by operators. The dynamic range of the pixel values may be truncated and the mean background level may vary from image to image, even across images of the same geographic area. Automated Target Recognition (ATR) algorithms that process these images require that pixel energy levels are normalized to remove systematic variations in amplitude that can cause these algorithms to fail. However, the adaptive nature of image normalization algorithms can introduce artifacts as well. In many cases the discarded original pixel values may contain valuable information. Calibrated SAS images, defined as images whose pixel values represent only the scattering properties of the seafloor or target, present a promising input to ATR because 1) they preserve amplitude information, which may provide stable feature measurements to ATR, and 2) they can be normalized without artifacts by taking into account propagation and scattering physics, potentially offering a superior method than current normalization techniques. For this grant, measured system parameter were used to form calibrated images from a high frequency synthetic aperture sonar system. Systems with large bandwidth and wide beams cannot directly estimate the scattering cross section, since that quantity is defined using a plane wave of a single frequency for the incident field. Correspondingly, the quantities produced by calibrated images must be understood as quantities averaged over the systems bandwidth hand beam width. Details of this method, and scattering cross section results are presented in the enclosed OCEANS2019 conference paper.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 30, 2019
- Accession Number
- AD1084798
Entities
People
- Daniel C Brown
- Derek R Olson
- J. T. Cobb
- Peter D. Romaine
Organizations
- Pennsylvania State University