ABSTRACT

In the process of using ground penetrating radar (GPR) to locate concealed power facilities such as underground cables, it is difficult to search the hyperbola generated by concealed facilities. To solve this problem, this paper proposes a two-step detection method to detect and locate the concealed facilities. First, the original echo image is pre-processed to enhance the weak signal and provide high-quality image for hyperbolic detection. And then, the first step of the two-step detection method is performed to extract the possible hyperbolic region. In this part, threshold processing method is used to reduce the clutter interference in the image, and the possible hyperbolic regions are detected and extracted based on the hyperbolic south open feature. In the second step, the trained faster Region-based Convolutional Neural Networks (faster-RCNN) is used to identify the extracted region, so as to obtain the position information of all hyperbolas in the echo image. By combining the hyperbolic detection results with the design drawings, the location of underground concealed facilities can be obtained. The proposed method is tested on the measured data, the experimental results show that the proposed method can accurately locate hyperbola from GPR data.

摘要

在使用探地雷达（GPR）定位地下电缆等隐蔽电力设施的过程中，很难搜索到隐蔽设施产生的双曲线。针对这一问题，本文提出了一种两步检测方法来检测和定位隐蔽设施。首先对原始回波图像进行预处理，增强弱信号，为双曲线检测提供高质量的图像。然后，执行两步检测方法的第一步以提取可能的双曲线区域。该部分采用阈值处理方法减少图像中的杂波干扰，并基于双曲南开特征检测和提取可能的双曲区域。第二步，使用训练好的faster-based Convolutional Neural Networks (faster-RCNN)对提取的区域进行识别，从而得到回波图像中所有双曲线的位置信息。通过将双曲线检测结果与设计图纸相结合，可以获得地下隐蔽设施的位置。在实测数据上对所提出的方法进行了测试，实验结果表明，所提出的方法能够准确地从探地雷达数据中定位双曲线。