Evaluating Operator Performance in Aided Airborne Mine Detection


In this paper we evaluate mine level detection performance of the human operator using high resolution mid-wave infrared (MWIR) imagery and compare it with the performance of automatic target recognition (ATR) like RX detector. Previous studies have shown that the anomaly detectors like the RX detector and even more sophisticated ATR techniques fall short of the performance achieved by human analyst for mine and minefield detection. There are three main objectives of the paper. First, we seek to establish performance bounds for mine detection using a single MWIR sensor under different conditions. Second, we evaluate the conditions under which the human visual system contributes significantly over and above RX anomaly detector. Third, we seek to qualitatively study the visual processes and mental models employed by the human operators to detect mines. A graphical user interface (HILgui) was developed using MATLAB to evaluate mine level detection performance for the operator. This interface is used to conduct a series of experiments examining performance for twenty subjects. The mine images varied systematically based on the time of day the images were collected, the type of terrain and type of mines. All the experiments were video-recorded and post-experiment interviews were conducted for qualitative analysis. Both qualitative and quantitative research techniques were used to gather and analyze the data. Results from different quantitative analysis including the accuracy of mine detection, propensity of false alarms and the time taken by the operator to mark individual targets are discussed. The mental models developed by the subjects for detection of mine targets are also discussed. Limitations of the current experiments and plans for future work are discussed. It is hoped that this systematic evaluation of a human operator in airborne mine detection will help in developing new and better ATR techniques and help identify critical features required in the operator interface for the warfighter-in-the-loop (WIL) minefield detection.


Electrical and Computer Engineering

Second Department

Business and Information Technology

Keywords and Phrases

Automatic Target Recognition (ATR); Mid-Wave Infrared (MWIR) Imagery; Mine Detection; RX Detector; Mines (Military explosives)--Detection

Document Type

Article - Journal

Document Version


File Type





© 2005 SPIE -- the International Society for Optical Engineering, All rights reserved.

Publication Date

01 Jan 2005