An Automatic GPR B-Scan Image Interpreting Model

作者：Zhou, Xiren; Chen, Huanhuan^*; Li, Jinlong

来源：IEEE Transactions on Geoscience and Remote Sensing, 2018, 56(6): 3398-3412.

DOI：10.1109/TGRS.2018.2799586

• Buried object detection
• data processing
• ground-penetrating radar (GPR)
• image processing

Ground-penetrating radar (GPR) has been widely used as a nondestructive tool for the investigation of the subsurface, but it is challenging to automatically process the generated GPR B-scan images. In this paper, an automatic GPR B-scan image interpreting model is proposed to interpret GPR B-scan images and estimate buried pipes, which consists of the preprocessing method, the open-scan clustering algorithm (OSCA), the parabolic fitting-based judgment (PFJ) method, and the restricted algebraic-distance-based fitting (RADF) algorithm. First, a thresholding method based on the gradient information transforms the B-scan image to the binary image, and the opening and closing operations remove discrete noisy points. Then, OSCA scans the preprocessed binary image progressively to identify the point clusters1 with downward-opening signatures, and PFJ further validates whether the point clusters with downwar-dopening signatures are hyperbolic. By utilizing OSCA and PFJ, point clusters with hyperbolic signatures could be classified and segmented from other regions even if there are some connections and intersections between them. Finally, the validated point clusters are fitted into the lower parts of hyperbolas by RADF that solves fitting problems with additional constraints related to the hyperbolic central axis. By integrating these methods, the proposed model is able to extract information from GPR B-scan images automatically and efficiently. The experiments on simulated and real-world data sets demonstrate the effectiveness of the proposed model.

• 出版日期2018-6
• 单位中国科学技术大学