基于Viola-Jones方法的探地雷达管线目标识别

Buried Pipes Recognition in GPR Data Based on Viola-Jones Algorithm

在城市地下管线的探测上,辨别探地雷达图像中的双曲线特征是管线探测的关键,但是实际情况中,城市的地下管线数量众多,专业人员短缺,自动化的探测识别手段的开发变得十分迫切。针对这一现状,对基于Viola-Jones方法的探地雷达目标识别手段进行了研究。在预处理过程中运用BMS(Background matrix subtraction)的方法生成背景矩阵,去除横向杂波干扰,然后采用Viola-Jones物体检测框架初步获取目标区域,最后使用C3(Column Connection Cluster)算法对目标区域二值化分割并筛选双曲线目标,得到的双曲线聚类,可用于管线的管径、深度获取。

In the detection of urban underground pipelines, it is the key to pipeline detection to distinguish the hyperbolic features in the ground penetrating radar image. However, in reality, the number of underground pipelines in the city is large, the shortage of professionals, and the development of automated detection and identification means becomes very urgent. As for this situation, the target detection method of ground penetrating radar based on Viola-Jones method is studied. In the preprocessing process, the background matrix is generated by the method of BMS(Background Matrix Subtraction) to remove the lateral clutter.Then the target area is obtained by the Viola-Jones object detection framework. Finally, the C3(Column Connection Cluster) algorithm is applied in the target area to get the hyperbolic features. The hyperbolic cluster can be used to calculate depth and radius of pipes in the future work.