Fast Prescreening for GPR Antipersonnel Mine Detection via Go Decomposition

• Published in: IEEE geoscience and remote sensing letters Vol. 16; no. 1; pp. 15 - 19
• Main Authors: Song, Xiaoji, Xiang, Deliang, Zhou, Kai, Su, Yi
• Format: Journal Article
• Language: English
• Published: Piscataway IEEE 01.01.2019; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Antipersonnel mine (APM) detection; Antipersonnel mines; Clutter; Decomposition; Detection; False alarms; Go Decomposition (GoDec); Ground penetrating radar; ground-penetrating radar (GPR); Image detection; Laboratory experiments; Land mines; Matrix decomposition; Methods; Mine detection; Numerical simulation; Object detection; prescreening; Principal components analysis; Probability theory; Radar; Robustness; Sparse matrices; Target detection; Target recognition
• ISSN: 1545-598X; 1558-0571
• DOI: 10.1109/LGRS.2018.2866331

Ground-penetrating radar (GPR) has been widely used for antipersonnel mine (APM) detection. However, its efficiency is often impaired by high false alarm rate (FAR) caused by the ground clutters. In this letter, a novel robust principal component analysis (RPCA)-based method is proposed for fast prescreening of APM in GPR image. Taking advantage of low rank and sparse structure of GPR image, the proposed method first adopts an efficient RPCA technique-Go Decomposition (GoDec)-to extract the target image. Then, thresholds are applied to the extracted image to detect the target and reject false alarms. The proposed method enjoys two advantages over traditional methods: 1) the ability of reducing FAR while maintaining high probability of detection (PD) in strong noise and clutter environment and 2) the fast detection guaranteed by the modified GoDec that yields results within several iterations. Extensive simulations and laboratory experiments are conducted to validate the proposed method, and the results are satisfactory (high PDs up to 99% and low FARs).