Recently non-destructive techniques for exploration and retrieve of underground information about different fields such as oil and gas exploration, geology, conduit and pipe location and archeology have been developed. Depending on kind of application, an appropriate sensor should be employed for mapping subsurface. For detection of buried objects at shallow depths, ground penetrating radar (GPR) is considered as the most suitable approach. This technique consists in the transmission and reception of electromagnetic waves by means of apparatus which has few meters exploration depth and resolution of several centimeters. The GPR device uses high frequency electromagnetic waves to acquire subsurface data. The device uses a transmitter antenna and closely spaced receiver antenna to detect changes in electromagnetic properties beneath them. The antennas are suspended just above the ground surface and the antennas are shielded to eliminate interference from sources other than directly beneath the device. The transmitting antenna emits a series of electromagnetic waves, which are distorted by differences in soil conductivity, dielectric permitivity, and magnetic permeability. The receiving antenna records the reflected waves for a specified length of time (in nanoseconds). The approximate depth of an object can be estimated with GPR, by adjusting for electromagnetic propagation conditions. The GPR signals that are captured by the receiving antenna are recorded in array of numerals, which can be converted to gray scale (or color) pixel values. The radargrams are essentially a vertical map of the radar reflection returned from subsurface objects. The shape of GPR radargrams obtained from cylindrical objects is in the form of semi-hyperbola. Interpretation of acquired GPR data needs an experienced geoscientist meaning high cost in terms of money and time. For detection of buried objects using GPR data limited works has been done. In current study two optimization methods namely the Genetic (GA) and Artificial Bee Colony (ABC) Algorithms has been applied for detecting and locating hyperbola signatures of small buried objects (mainly pipes and channels). In the first step, two data sets including synthetic data using GprMax2d program and real data surveyed in the of Isfahan University of Technology (IUT) campus were constructed. Behaviour of synthetic data shows that changing the depth and radius of buried object, changes the shape of the semi-hyperbola showing the valuable geometrical information hiden in GPR responses. In the next step the data has been preprocessed using following operations: 1) Dewow filtering and removing DC bias 2) Background removal 3) Manual Gain function and 4) binarization. The above mentioned algorithms were applied to real and synthetic data for extracting features and estimation depth and radius of buried cylindrical objects (mainly metallic pipes). Results show that GA and ABC both are efficient optimization algorithms and can be used in acquiring geometrical patterns hiden in GPR images. Finally it is concluded that both ABC and GA have acceptable performances in detecting geometric parameters but the employed ABC algorithm outperforms the GA results owing mainly to its speed, robustness and accuracy. In spite of the fact that the associated errors in estimating geometric parameters out of real data compared with that of synthetic data is some what greater but the obtained results have enough accuracy and precision in practical works and can be used as an indicator and guide for locating buried objects.