Random Neural Network Filter for Land Mine Detection

The two primary measures of land mine detection performance are the probability of detection Pd and the probability of false alarm Pfa.These two measures are highly interdependent and must be evaluated together. The relationship between the two probabilities directly a ects the overall performance of the sensor in the eld. In this paper we introduce a novel false alarm non-parametric lter based on the Random Neural Network (RNN) model [2, 3, 4] and the -Technique [1], the study is based on mine detection using Electromagnetic Induction (EMI) sensors. The mine eld data are pre-processed via the -Technique before applying it to the RNN. The RNN has a prede ned structure that tries to implement a mapping close enough in some precise sense to the discrimination function between non-mine and mine patterns [8]. Limited number of non-mine and mine patterns, extracted from a small calibration area for a certain mine eld provided by DARPA [5], are used for training the RNN. We show that the RNN gives e ective decisions on patterns measured on other locations using di erent EMI sensor. The results show that the RNN produces probability of detection up to 100 percent with a substantial reduction of false alarms over the -Technique (up to 40 percent false alarm ltering). Key-words: Mine Detection, Binary Random Neural Network, -Technique, False Alram Filtering, Anthropic Clutter, Decision Boundary Function Approximation, EMI sensor.