This paper presents a comparative study of two algorithms for detecting and analyzing the characteristic shapes of reflection obtained as a result of Ground-Penetrating Radar (GPR) scanning technology. The first algorithm is a sub-array processing method that uses direction-of-arrival algorithms and the matched filter technique; this approach is implemented in SPOT-GPR (release 1.0), a new freeware tool for the detection and localization of targets in radargrams. The second algorithm, APEX, is based on machine learning and pattern recognition techniques and it allows finding the coordinates of apexes and further characteristic points of hyperbolas in radargrams. Both software solutions are implemented in MATLAB environment. As a first step, we compare the accuracy of our algorithms when applied to synthetic data, calculated by using the open-source finite-difference time-domain simulator gprMax; the scenarios are two concrete cells hosting different metallic and dielectric targets. Then, we compare the accuracy of our algorithms when applied to experimental data, recorded over district heating pipes in a trench, with known geometry and depth of the pipes. For the latter scenario, we have also generated a gprMax radargram, matching the geometry and scanning settings of the real one; both algorithms are tested on this synthetic radargram, as well. Overall, both algorithms perform well and rather uniformly in localizing the targets. The accuracy of the algorithms is at centimeter level, which is sufficient in most applications.