3D Segmentation of ground penetrating radar data for landmine detection

In this paper, we propose a 3D image analysis method for landmine detection from vehicle mounted Ground Penetrating Radar (GPR) data, using adaptive filtering technique and image segmentation algorithm. Conventional methods for landmine detection seldom exploit the 3D information (C-Scan) and the discrimination algorithms used are computationally expensive. In our method, we have used 3D adaptive filtering technique (3D LMS), which is suitable for real-time applications for locating the anomalies in C-Scan data. Gradient Vector Flow Deformable Model based segmentation algorithm is applied on the localized regions to obtain a 3D volumetric segmented image. The segmented regions' shape features are used to further discriminate targets from clutters to obtain a 3D segmented image of landmines. Data collected from 3D Radar system and Vector Network Analyzer (VNA) based SFCW-GPR is used for performance evaluation of the proposed algorithm. Our proposed method has shown better performance than other state-of-the-art algorithms for landmine detection, in terms of its Receiver Operating Characteristic (ROC) measure.