A Ground-penetrating Radar Object Detection Method Based on Deep Learning

The Ground-penetrating radar (GPR) is widely applied in the detection tasks. This paper proposes a new detection method of the GPR objects, which is based on the Cascade Regional Convolutional Neural Network (Cascade R-CNN). The proposed method effectively addresses the problems caused by the existing methods with the high time cost, low detection accuracy and poor adaptability. Additionally, an adaptive clutter filter algorithm is also proposed to realize the operation of data preprocessing when constructing the data set, so as to increase the signal-to-noise ratio. The experiment results on the collected and simulation data show that the proposed method has the great performance, achieving the average precision more than 85%. It accurately detects the buried objects in different environments, demonstrating the good generalization and robustness of the proposed method.