SBIR-STTR Award

For this Phase I effort, Observera proposes to develop an attribute based system for automated image quality assessment of Synthetic Aperture Radar (SAR) Imagery. Observera's expertise in SAR image analysis, image quality assessment and digital image processing algorithms are uniquely suited to this program. The imagery community uses the Radar National Imagery Interpretation Rating System (RNIIRS) as a metric to describe SAR image quality as a function of image utility in task performance. However, RNIIRS is derived from subjective, scene dependent criteria ill suited to automation. Observera personnel have participated in recent studies have correlated RNIIRS with the image domain attributes of brightness, focus, contrast and noise. We will demonstrate that these attributes can be measured algorithmically in formed digital SAR imagery. The metrics from these algorithms will be fed into a classifier that analyzes the metrics and outputs an image quality score that can be correlated back to an RNIIRS rating. This measures the potential usability of the imagery with respect to predicted quality based on collection parameters. Observera will deliver image quality algorithms to measure the attributes of brightness, contrast, focus and noise, as well as a classifier to determine quality based on analysis of these measurements.

Observera proposes to build on the success of our Phase I SBIR "Automated synthetic Aperture Radar Image Quality Assessment" by developing an operational prototype system that eliminates Radar national Imagery Interpretability Scale (RNIITS) rating based on digitally measured attributes. The projected abundance of SAR imagery will soon overwhelm the capacity for human quality review, mandating automated assessment. RNIIRS is the accepted quality yardstick, but it is a subjective, scene dependent metric ill suited to automation. Building on past research, Observera established that RNIIRS is correlated with the image domain attributes of brightness, focus, noise, and contrast which correlated with the image domain attributes of brightness, focus, noise, and contrast which can be measured automatically in formed SAR imagery. The digitally measured attributes are used as coefficients to automatically derive an estimate of RNIIRS. Observera will further refine and integrate the algorithms and estimation techniques developed in Phase I into a prototype system that will automatically estimate RNIIRS in real-time from formed UAV collected SAR imagery. Furthermore, Observera will demonstrate that this methodology can be used to extract metrics that can determine the suitability of an image for ATR processing in the Moving and Stationary Target Acquisition and recognition (MSTAR) system to lower false alarm rates caused by poor quality imagery.