Fuzzy clustering for land mine detection

Fuzzy Clustering is applied to the problem of detecting landmines. Sensor data generated by a Ground Penetrating Radar (GPR) is processed to detect the mines. The GPR produces a three-dimensional array of intensity values, representing a volume below the surface of the ground. Features are computed from this array and clustered using a fuzzy Competitive Agglomerative (CA) algorithm. Prototypes are produced by the clustering algorithm and used to detect landmines. A novel aspect of this work is that the prototypes are not used in a nearest-prototype style classifier, which would be the standard approach. Rather, the prototypes are used to provide a reliable indicator of the strength and pattern of a return at a location beneath the surface. Results on real, difficult data are provided that indicate that the fuzzy clustering produces more reliable detection outputs. In particular, the false alarm rates are much lower than those of the existing system.