Abstract: Ground Penetrating Radar (GPR) represents a widely employed nondestructive testing technique in the realm of water supply pipeline leakage detection. However, currently, when GPR is utilized for water - supply pipeline leakage detection, it primarily relies on analyzing characteristics such as the energy, velocity, and frequency of GPR signals to identify the presence of leakage, making it challenging to directly determine the pipeline’s leakage direction and area. To address this, this paper puts forward a GPR - based underground water - supply pipeline leakage identification method grounded in reverse - time migration imaging. This method employs threshold segmentation to extract the slope of the hyperbolic diffraction - wave tail in GPR data related to water - supply pipeline leakage and integrates this slope with the conversion formula for the electromagnetic wave velocity of the subsurface medium to deduce the electromagnetic - wave velocity of the subsurface medium, thereby obtaining a more precise migration velocity. Given that the calculated migration velocity corresponds to the electromagnetic - wave velocity of the medium in the non - leakage region, and the electromagnetic - wave velocity in the leakage region is lower than that in the non - leakage region, performing reverse - time migration imaging on GPR data of water - supply pipeline leakage with this migration velocity can better converge the hyperbolic diffraction generated by the pipeline. Meanwhile, the diffraction waves from the leakage region are not fully converged, enabling high - accuracy identification of the leakage region and leakage direction. The numerical test results demonstrate that in the case of upper - side leakage, the pipeline echo shifts downward more significantly or even vanishes due to pipeline submersion, and the echo hyperbola in the leakage area basically does not converge. For left - and right - side leakage, the pipeline echo hyperbola converges well whereas the leakage - area echo hyperbola converges poorly. In the case of lower - side leakage, both echo hyperbolas converge well. These characteristics can serve as a basis for distinguishing between leakage directions and leakage areas.