Underground Cylindrical Objects Detection and Diameter Identification in GPR B-Scans via the CNN-LSTM Framework

• DOI: 10.3390/electronics9111804
• Publication Date: 2020-11-01T01:39:56Z
• Authors: Wentai Lei; Jiabin Luo; Feifei Hou; Long Xu; Ruiqing Wang; Xinyue Jiang
• Keywords: Ground-Penetrating Radar; Identification; Hyperbola; Transfer of learning

Ground penetrating radar (GPR), as a non-invasive instrument, has been widely used in the civil field. The interpretation of GPR data plays vital role underground infrastructures to transfer raw interested information, such diameter. However, diameter identification objects B-scans is tedious and labor-intensive task, which limits further application field environment. paper proposes deep learning-based scheme solve issue. First, an adaptive target region detection (ATRD) algorithm proposed extract regions from that contain hyperbolic signatures. Then, Convolutional Neural Network-Long Short-Term Memory (CNN-LSTM) framework developed integrates Network (CNN) Long (LSTM) network hyperbola features. It transfers task into classification. Experimental results conducted on both simulated datasets demonstrate promising performance for identification. CNN-LSTM achieves accuracy 99.5% 92.5% datasets.