PrintPrint

Awards

Topic Information Award/Contract Number Proposal Information Company Performance
Period
Award/Contract
Value
Abstract

H-SB012.2-003
Objective, Quantitative Image Quality Measurements and Metrics for Screener Imaging Technologies

HSHQDC-13-C-00073 DHS SBIR-2012.2-H-SB012.2-003-0001-II
(DHS SBIR-2012.2 Phase II)
Objective X-ray Image Display Evaluation (OXIDE)

Charles River Analytics Inc.
625 Mount Auburn Street
Cambridge, MA 02138-4555

09/03/2013
to
09/02/2015
$749,394.67

Transportation Security Officers (TSOs) are tasked with exploiting X-ray inspection systems to detect potential threats. To ensure efficient operation and the safety of the traveling public, TSOs must maintain a 100% probability of detection (Pd) while minimizing screening time. To achieve these objectives, X-ray inspection systems must operate in accordance with manufacturer specifications, and be calibrated to maximize Pd and reduce screening time. The ASTM X-ray Test Object aims to evaluate image quality (IQ) with respect to Pd to support optimal system calibration. However, the ASTM is fundamentally flawed for several reasons: (1) it is prone to operator bias; (2) it is not directly representative of real-world operation; (3) it does not quantify the relationship between IQ and Pd; and (4) it cannot handle moving objects, and is therefore ineffective for exploring the use of continuously rotating conveyor belts to speed up the screening process. To address these concerns, we designed, developed, and demonstrated a prototype for Objective X-ray Image Display Evaluation (OXIDE). OXIDE uses a predictive approach to assess functional IQ during normal operation, producing a single General Image Score that can cue the operator to potential image degradations. The system leverages our existing object detection and IQ technologies to achieve robust and purely objective IQ evaluation for X-ray screening systems. The Phase I prototype accurately predicts functional IQ on real X-ray imagery that has been randomly degraded, corresponding well to a human's qualitative interpretation of relative image quality, and demonstrating the feasibility of our predictive IQ assessment approach.