This paper describes an algorithm for land mine detection in GPR data that uses edge histogram descriptors
for feature extraction and fuzzy K-Nearest Neighbors (K-NN) for confidence assignment. First, an LMS algorithm
for anomaly detection is used to focus attention and identify candidate signatures that resemble mines.
Second, translation invariant features are extracted based on spatial distribution of edges in the 3-D GPR
signatures. Specifically, each 3-D signature is divided into sub-signatures, and the local edge distribution
for each sub-signature is represented by a histogram. To generate the histogram, local edges are categorized
into five types: vertical, horizontal, diagonal, anti-diagonal, and non-edges. Next, the training signatures
are clustered to identify prototypes. The main idea is to identify few prototypes that can capture the
variations of the signatures within each class. These variations could be due to different mine types,
different soil conditions, different weather conditions, etc. Fuzzy memberships are assigned to these
representatives to capture their degree of sharing among the mines and false alarm classes.
Finally, fuzzy K-NN based rules are used to assign a confidence value to distinguish true detections from
false alarms. The proposed algorithm is applied to data acquired from three
outdoor test sites at different geographic locations.