Decision fusion for improved detection of buried radioactive objects

Detection of buried radioactive objects faces challenges such as low energy counts and strong background clutters due to the burial of the targets. Classical detection methods such as the constrained energy minimization (CEM) and the RX method, when applied separately, may not be able to yield satisfactory results. In this paper, we propose to combine detection results from individual detectors through decision fusion, so that improved detection performance may be achieved. For a given data set, different detection methods are employed to yield binary decisions (“radiation” or “non-radiation”) separately. Then the decisions are fused using majority voting. In addition, the scores generated from each detector (before yielding binary decisions) are also utilized to assist the fusion process. The proposed method is demonstrated to improve over the individual detectors using survey data collected on simulation targets.