Object Detection and Localization in GPR B- Scan images using Hyperbola Fitting Algorithm

Abstract

Ground-penetrating radar [GPR] is a non-invasive method, that generates an electromagnetic wave to map a subsurface object. It finds applications in various fields such as civil / structural engineering, environmental, Geotechnical and Military applications. This research is conducted at the Civil Engineering laboratory, IISc, Bangalore to collect the experimental data, using the Mala Pro-ex ground-coupled antenna GPR system from Mala Geoscience. GPR technology emits electromagnetic pulses into the ground and captures the reflected echoes from the target,enabling a detailed visualization of the subsurface targets.This paper presents the Hyperbola fitting algorithm for accurate detection and localization of the underground objects such as metallic and nonmetallic. It involves several steps, including preprocessing, identifying hyperbolas, fitting hyperbolas, determining target locations and depths, and validating the results. Following the fitting of the hyperbola, the algorithm proceeds to estimate the target's position and depth using the hyperbola's parameters. These estimated values for the target location and depth are subsequently validated through a model generated using the gprMax numerical modeling tool and is also tested on experimental data. There are numerous iterations of hyperbola fitting algorithms, and their implementation varies according to the attributes of the GPR data and the subject of interest. This method provides significant advantages for engineers before excavation of the ground. The algorithm has been developed and implemented using MATLAB. It has been rigorously tested using both experimental data and simulated data generated through gprMax Numerical modeling tool. The algorithm achieves a minimum error of 0.02 in the lateral position (x) and 0.01 in the depth (Y) direction. These results demonstrate the algorithm's accuracy and reliability in identifying and locating buried objects.