Buried wire detection using ground penetrating radars


Tezin Türü: Yüksek Lisans

Tezin Yürütüldüğü Kurum: Orta Doğu Teknik Üniversitesi, Mühendislik Fakültesi, Elektrik ve Elektronik Mühendisliği Bölümü, Türkiye

Tezin Onay Tarihi: 2017

Öğrenci: UTKU YILMAZ

Danışman: GÖZDE AKAR

Özet:

Buried explosives pose great threat for national security of the countries struggling with these explosives as the damage caused by them increase day by day. There are several sensors developed for detection of buried explosives such as land mines, unexploded ordnances and improvised explosive devices (IEDs). Among these, IEDs are quite hard to detect with majority of sensors due to their irregular shape and contents. The difficulties in IED detection have led researchers to aim the triggering mechanisms of IEDs. As the new jamming systems in the military industry can successfully block the wireless control links of IEDs, the threat has shifted to the use of command wires. So, the detection of buried command wires become a critical ability for buried explosive detection systems. Ground penetrating radars have shown their capabilities on detection of buried objects in many operational concepts. Ground penetrating radars can construct the 3-D image of the subsurface medium with high spatial and temporal resolution and distinguish objects with different electromagnetic properties. In this thesis, wire detection problem is studied using ground penetrating radars. Firstly, extensive simulations are carried out on gprMax software, a powerful open source FDTD simulation environment, by changing simulation parameters such as transmitting frequency, electromagnetic properties of soil and clutter, depth and radius of wire, in order to observe the effects of these parameters. Then, a simulated 3-D GPR database is generated consisting of wires in different orientations together with different types of clutter. In the second step, wire detection and classification problem is studied. The possible wire locations are found using a morphologically improved version of 2-D LMS filtering. Then a novel 3-D feature set is extracted with the help of 3-D curve reconstruction algorithm. Using the generated database, SVM classifier is trained and the performance of proposed algorithms is shown.