This thesis is concerned with detection logarithms of non-metallic anti-personnel (AP) land mines and the related signal processing using stepped-frequency ground penetrating radar (SF-GPR) technique. Modern land mines are essentially made out of plastic and ceramic materials. This makes their detection using GPR sensor almost impossible without proper signal processing. Many signal processing algorithms have been presented in the literature and successfully applied to GPR data for the detection of AP land mines. Therefore there is a need to compare these methods as regarding efficiency and hard- and software requirements. The thesis presents most common signal processing techniques used for SF-GPR based detection of buried objects. These techniques have been investigated, implemented and compared to each other as regarding their ability to separate the land mine and noise signals. The algorithm that performed best in these comparison is called Independent Component Analysis algorithm, which has demonstrated the ability to eliminate the GPR clutter and extract the target signal. Furthermore, combining the wavelet packet transform with the higher-order-statistics has shown to be very effective in the GPR signal processing. All experimental results presented in the thesis are based on real measured data obtained from an experimental SF-GPR system. The system has been developed and built at the Institute of Electronics, Signal Processing and Communications Engineering (IESK), University of Magdeburg, Magdeburg, Germany.