Anti-personnel Mine Detection by Sparse Representation of GPR B-scan Radargram Image

Ground penetrating radar is an efficient close-range remote sensing and non-destructive-testing technique, which probes ultra-wide-band electromagnetic pulses to detect underground targets. Then theirs geometric features and material property characters will been figured out. A major clutter interference coupling with the response of the subsurface-buried-landmine is direct air wave and direct ground reflected waves, which hinder shallow buried landmine precise detection and discrimination. As for shallow buried targets, in this paper, a clutter and target response echo decoupling method with Linear-Gain-Compensation-Robust-PCA algorithm is proposed based on low-rank prior of clutters and target distribution spatial sparse prior. Compared with Principal Component Analysis, mean subtract algorithm, the performance of LGC-RPCA method is tested with simulations and indoor sand-box experiments. These demonstrate that the LGC-RPCA method with inexact augmented Lagrangian function is an efficient algorithm for shallow buried landmine detection and discrimination. Then, it will reduce the false alarm rate for identification of shallow buried anti-personnel Mine.