Automatic Landmine Detection by Pattern Trace

This paper presents a novel approach for auto detection of Landmine in Ground Penetrating Radar (GPR) data using a machine learning technique. GPR scan data has noise and clutter due to different permittivity of material and different scattering effect of electromagnetic reflections caused by rocks, buried materials etc. The proposed approach includes image processing methods such as normalization, background noise removal, image smoothing, edge detection with probability based threshold and clustering algorithm for feature extraction and pattern tracing technique to detect hyperbola pattern. Classification of landmine from object has been done from result that fits in a range that is derived from various experimental results for auto detection of landmine.