A spatial display for Ground-Penetrating Radar change detection

Abstract

Ground-Penetrating Radar (GPR) enables the exploration and mapping of subterranean volumes for applications such as construction, humanitarian demining, archeology, and environmental science. In each of these applications, special signal processing pipelines have been developed to reduce noise and reject clutter for optimal object detection and tracking. Change Detection (CD) is one approach to solving these signal detection challenges by leveraging the concept that changes are more relevant than absolute measurements. This research focuses on the Gopher vehicle-mounted CD GPR system. Regardless of the application, GPR data must be interpreted by some intelligence, whether human or artificial. Traditional GPR interfaces present the raw GPR data to an operator in cross sections organized by time and depth. The intent of these displays is to allow a human operator to formulate a mental model and plan of action. After a human factors evaluation, this presentation was identified as suboptimal, and a new display was designed to present GPR data. The new display organizes data in a spatial manner and presents the information to the operator on a map. The display was tested in a human subjects experiment with thirty untrained volunteers and two expert operators measuring the signal detection properties of the display compared with a traditional temporally-organized display. The display and operator system was evaluated using signal detection theory analysis. The new spatial display was quantitatively superior as evidenced by a 4.7% increase in correctness of the subjects' classifications and a 29% decrease in miss percentage. Qualitatively, 83% of subjects preferred the new interface and 7% had no preference. The collective intelligence implications of this system were investigated by simulating voting committees of operators. Committee performance was superior to expert operators and to top performers in several respects.