Ground penetrating radar target detection method based on optimized faster RCNN and k-means clustering
Changle Xin, Wentai Lei, Chaopeng Luo, Wei Chang Xue · IET conference proceedings. · 2024
Ground Penetrating Radar (GPR) technology is a widely used detection technology for underground target objects, which can perform non-destructive detection of underground targets. Due to the complexity of the underground environment and the interference of clutter, manual processing of GPR data is a time-consuming and laborious task. Therefore, an accurate and fast automatic underground target detection method is urgently needed. We propose a method that can achieve fast goals in complex multi-goal situations. The simplified ResNet-50 network is used to detect dense intersecting hyperbolic curves and reduce the computation of the network. At the same time, we use k-means clustering algorithm to optimize the anchor in faster-Rcnn based on the characteristics of GPR hyperbola. In addition, this method is also suitable for target detection on GPR images under random media.