Publications

Abstract not provided.

Single Image SICD-Based Automatic Object Processing (SIS-AOP) is an automatic object identification tool for SAR imagery. It ingests a SAR image in standard SICD format, and it will run a suite of algorithms to cue possible vehicle detections, cull those detections and then ultimately label them either as detections only or possible expound to give a class-level ID or a vehicle-type ID. The SIS-AOP results are given in an XML (Extensible Markup Language) output format. This document defines the elements in the SISAOPR XML output format.

For machine vision, one of the most important operations is fast and effective object cueing or segmentation. Sandia National Labs has a long history of development and implementation of very fast and effective cueing/segmentation algorithms. This report covers the history, motivation and implementation of evolving frameworks (Sandia FOA Frameworks) upon which this long legacy of successful algorithms are built. The report describes the innovative microprocessor implementation, enabling extremely fast morphological processing, combined with a novel adaptive quantization front - end and a feature - based backend that resulted in Sandia developing fast and effective cueing in a wide variety of applications, from defect detection to SAR ATR. The report covers evolution from Sandia FOA 1.0 Framework (1995) to current Sandia FOA 4.0 Framework (2021). Requirements for the cueing algorithm for SIS - AOP (FOA_SIS - AOP) that drove the Sandia FOA 4.0 Framework development are discussed, along with information on how to use the Sandia FOA Frameworks.

Phenomenology-Informed (PI) Machine Learning is introduced to address the unique challenges faced when applying modern machine-learning object recognition techniques to the SAR domain. PI-ML includes a collection of data normalization and augmentation techniques inspired by successful SAR ATR algorithms designed to bridge the gap between simulated and real-world SAR data for use in training Convolutional Neural Networks (CNNs) that perform well in the low-noise, feature-dense space of camera-based imagery. The efficacy of PI-ML will be evaluated using ResNet, EfficientNet, and other networks, using both traditional training techniques and all-SAR transfer learning.

Phenomenology-Informed (PI) Machine Learning is introduced to address the unique challenges faced when applying modern machine-learning object recognition techniques to the SAR domain. PI-ML includes a collection of data normalization and augmentation techniques inspired by successful SAR ATR algorithms designed to bridge the gap between simulated and real-world SAR data for use in training Convolutional Neural Networks (CNNs) that perform well in the low-noise, feature-dense space of camera-based imagery. The efficacy of PI-ML will be evaluated using ResNet, EfficientNet, and other networks, using both traditional training techniques and all-SAR transfer learning.

Automatic or assisted target recognition (ATR) is an important application of synthetic aperture radar (SAR). Most ATR researchers have focused on the core problem of declaration-that is, detection and identification of targets of interest within a SAR image. For ATR declarations to be of maximum value to an image analyst, however, it is essential that each declaration be accompanied by a reliability estimate or confidence metric. Unfortunately, the need for a clear and informative confidence metric for ATR has generally been overlooked or ignored. We propose a framework and methodology for evaluating the confidence in an ATR system's declarations and competing target hypotheses. Our proposed confidence metric is intuitive, informative, and applicable to a broad class of ATRs. We demonstrate that seemingly similar ATRs may differ fundamentally in the ability-or inability-to identify targets with high confidence.

The INTEROP Achievement Award will be given to those customer organizations that make the most effective use of internetworking technology to further their own specific business aims. This paper is an application for this award by Sandia National Laboratories. Given are the network application, topology, and the types of systems to which it is applied.(JEF)