基于自动边缘检测和模板匹配的探地雷达目标检测方法

Target Detection Based on Automatic Threshold Edge Detection and Template Matching Algorithm in GPR

摘要: 探地雷达对地下复杂场景中的目标进行探测时，由于地下介质的非均匀，导致回波信号在传播路径和传播衰减上的扰动，造成探地雷达记录剖面中双曲线特征的畸变、错位、缺失等现象，传统的基于双曲线顶点检测的目标定位方法不再有效。本文提出了一种基于自动阈值边缘检测和模板匹配的散射曲线顶点检测方法来抑制双曲线特征形变和噪声的影响，提高顶点检测的精度。该方法通过探地雷达记录剖面沿测线维的一维能量曲线确定目标个数的预设区间；然后基于自动匹配阈值的方式对探地雷达记录剖面进行边缘检测，得出估计的顶点坐标；再基于估计的顶点坐标建立模板并进行模板匹配，得出匹配后的目标顶点；最后对目标顶点采用聚类分析的方法进行过滤，去除虚假顶点，得出探地雷达记录剖面中的目标顶点的检测结果。通过对仿真和实测数据的处理，结果表明本方法能有效的抑制双曲特征形变和噪声的影响，具有更高的检测精度。

Abstract: When using ground-penetrating radar (GPR) detects targets in complex underground scene, due to the non-uniformity of underground medium, the disturbance of echo signal at the aspects of propagation path and propagation attenuation leads to the phenomenon of such as dislocation, missing, distortion characteristic of scattering curve in GPR record profile. Traditional methods of target localization based on hyperbolic vertex detection are no longer valid. In this paper, a method based on automatic threshold edge detection and template matching was proposed to reduce the influence of hyperbolic distortion and noise and improved the accuracy of vertex detection. Firstly, one-dimensional energy curves along the lateral line are recorded by GPR to determine preset number of targets. A preset range of target number was determined by recording a one-dimensional energy curve along a survey line by the GPR record profile. Secondly, the edge of GPR record profile was detected based on automatic matching threshold to obtain estimated vertex coordinates. Based on the estimated vertex coordinates, templates were established and template matching process was carried out to obtain matching target vertex; finally, target vertices were filtered by clustering analysis and false vertices were removed, and detection results of target vertex in GPR profile were obtained. Simulation and on-site GPR profile processing results show that the proposed method can effectively suppress the effects of hyperbolic deformation and noise with higher detection accuracy.